Skip to content
This repository
Fetching contributors…

Cannot retrieve contributors at this time

file 3566 lines (2816 sloc) 115.385 kb
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566
import types
from functools import wraps
import numpy as np
import datetime
import collections

from pandas.compat import(
    zip, builtins, range, long, lzip,
    OrderedDict, callable
)
from pandas import compat

from pandas.core.base import PandasObject
from pandas.core.categorical import Categorical
from pandas.core.frame import DataFrame
from pandas.core.generic import NDFrame
from pandas.core.index import Index, MultiIndex, _ensure_index, _union_indexes
from pandas.core.internals import BlockManager, make_block
from pandas.core.series import Series
from pandas.core.panel import Panel
from pandas.util.decorators import cache_readonly, Appender
import pandas.core.algorithms as algos
import pandas.core.common as com
from pandas.core.common import(_possibly_downcast_to_dtype, isnull,
                               notnull, _DATELIKE_DTYPES, is_numeric_dtype,
                               is_timedelta64_dtype, is_datetime64_dtype)

from pandas import _np_version_under1p7
import pandas.lib as lib
from pandas.lib import Timestamp
import pandas.tslib as tslib
import pandas.algos as _algos
import pandas.hashtable as _hash

_agg_doc = """Aggregate using input function or dict of {column -> function}

Parameters
----------
arg : function or dict
Function to use for aggregating groups. If a function, must either
work when passed a DataFrame or when passed to DataFrame.apply. If
passed a dict, the keys must be DataFrame column names.

Notes
-----
Numpy functions mean/median/prod/sum/std/var are special cased so the
default behavior is applying the function along axis=0
(e.g., np.mean(arr_2d, axis=0)) as opposed to
mimicking the default Numpy behavior (e.g., np.mean(arr_2d)).

Returns
-------
aggregated : DataFrame
"""


# special case to prevent duplicate plots when catching exceptions when
# forwarding methods from NDFrames
_plotting_methods = frozenset(['plot', 'boxplot', 'hist'])

_common_apply_whitelist = frozenset([
    'last', 'first',
    'head', 'tail', 'median',
    'mean', 'sum', 'min', 'max',
    'cumsum', 'cumprod', 'cummin', 'cummax', 'cumcount',
    'resample',
    'describe',
    'rank', 'quantile', 'count',
    'fillna',
    'mad',
    'any', 'all',
    'irow', 'take',
    'idxmax', 'idxmin',
    'shift', 'tshift',
    'ffill', 'bfill',
    'pct_change', 'skew',
    'corr', 'cov', 'diff',
]) | _plotting_methods

_series_apply_whitelist = \
    (_common_apply_whitelist - set(['boxplot'])) | \
    frozenset(['dtype', 'value_counts', 'unique', 'nunique',
               'nlargest', 'nsmallest'])

_dataframe_apply_whitelist = \
    _common_apply_whitelist | frozenset(['dtypes', 'corrwith'])


class GroupByError(Exception):
    pass


class DataError(GroupByError):
    pass


class SpecificationError(GroupByError):
    pass


def _groupby_function(name, alias, npfunc, numeric_only=True,
                      _convert=False):
    def f(self):
        self._set_selection_from_grouper()
        try:
            return self._cython_agg_general(alias, numeric_only=numeric_only)
        except AssertionError as e:
            raise SpecificationError(str(e))
        except Exception:
            result = self.aggregate(lambda x: npfunc(x, axis=self.axis))
            if _convert:
                result = result.convert_objects()
            return result

    f.__doc__ = "Compute %s of group values" % name
    f.__name__ = name

    return f


def _first_compat(x, axis=0):
    def _first(x):
        x = np.asarray(x)
        x = x[notnull(x)]
        if len(x) == 0:
            return np.nan
        return x[0]

    if isinstance(x, DataFrame):
        return x.apply(_first, axis=axis)
    else:
        return _first(x)


def _last_compat(x, axis=0):
    def _last(x):
        x = np.asarray(x)
        x = x[notnull(x)]
        if len(x) == 0:
            return np.nan
        return x[-1]

    if isinstance(x, DataFrame):
        return x.apply(_last, axis=axis)
    else:
        return _last(x)


def _count_compat(x, axis=0):
    return x.size


class Grouper(object):
    """
A Grouper allows the user to specify a groupby instruction for a target object

This specification will select a column via the key parameter, or if the level and/or
axis parameters are given, a level of the index of the target object.

These are local specifications and will override 'global' settings, that is the parameters
axis and level which are passed to the groupby itself.

Parameters
----------
key : string, defaults to None
groupby key, which selects the grouping column of the target
level : name/number, defaults to None
the level for the target index
freq : string / freqency object, defaults to None
This will groupby the specified frequency if the target selection (via key or level) is
a datetime-like object
axis : number/name of the axis, defaults to None
sort : boolean, default to False
whether to sort the resulting labels

additional kwargs to control time-like groupers (when freq is passed)

closed : closed end of interval; left or right
label : interval boundary to use for labeling; left or right
convention : {'start', 'end', 'e', 's'}
If grouper is PeriodIndex

Returns
-------
A specification for a groupby instruction

Examples
--------
>>> df.groupby(Grouper(key='A')) : syntatic sugar for df.groupby('A')
>>> df.groupby(Grouper(key='date',freq='60s')) : specify a resample on the column 'date'
>>> df.groupby(Grouper(level='date',freq='60s',axis=1)) :
specify a resample on the level 'date' on the columns axis with a frequency of 60s

"""

    def __new__(cls, *args, **kwargs):
        if kwargs.get('freq') is not None:
            from pandas.tseries.resample import TimeGrouper
            cls = TimeGrouper
        return super(Grouper, cls).__new__(cls)

    def __init__(self, key=None, level=None, freq=None, axis=None, sort=False):
        self.key=key
        self.level=level
        self.freq=freq
        self.axis=axis
        self.sort=sort

        self.grouper=None
        self.obj=None
        self.indexer=None
        self.binner=None
        self.grouper=None

    @property
    def ax(self):
        return self.grouper

    def _get_grouper(self, obj):

        """
Parameters
----------
obj : the subject object

Returns
-------
a tuple of binner, grouper, obj (possibly sorted)
"""

        self._set_grouper(obj)
        return self.binner, self.grouper, self.obj

    def _set_grouper(self, obj, sort=False):
        """
given an object and the specifcations, setup the internal grouper for this particular specification

Parameters
----------
obj : the subject object

"""

        if self.key is not None and self.level is not None:
            raise ValueError("The Grouper cannot specify both a key and a level!")

        # the key must be a valid info item
        if self.key is not None:
            key = self.key
            if key not in obj._info_axis:
                raise KeyError("The grouper name {0} is not found".format(key))
            ax = Index(obj[key],name=key)

        else:
            ax = obj._get_axis(self.axis)
            if self.level is not None:
                level = self.level

                # if a level is given it must be a mi level or
                # equivalent to the axis name
                if isinstance(ax, MultiIndex):

                    if isinstance(level, compat.string_types):
                        if obj.index.name != level:
                            raise ValueError('level name %s is not the name of the '
                                             'index' % level)
                    elif level > 0:
                        raise ValueError('level > 0 only valid with MultiIndex')
                    ax = Index(ax.get_level_values(level), name=level)

                else:
                    if not (level == 0 or level == ax.name):
                        raise ValueError("The grouper level {0} is not valid".format(level))

        # possibly sort
        if (self.sort or sort) and not ax.is_monotonic:
            indexer = self.indexer = ax.argsort(kind='quicksort')
            ax = ax.take(indexer)
            obj = obj.take(indexer, axis=self.axis, convert=False, is_copy=False)

        self.obj = obj
        self.grouper = ax
        return self.grouper

    def _get_binner_for_grouping(self, obj):
        raise NotImplementedError

    @property
    def groups(self):
        return self.grouper.groups

class GroupBy(PandasObject):

    """
Class for grouping and aggregating relational data. See aggregate,
transform, and apply functions on this object.

It's easiest to use obj.groupby(...) to use GroupBy, but you can also do:

::

grouped = groupby(obj, ...)

Parameters
----------
obj : pandas object
axis : int, default 0
level : int, default None
Level of MultiIndex
groupings : list of Grouping objects
Most users should ignore this
exclusions : array-like, optional
List of columns to exclude
name : string
Most users should ignore this

Notes
-----
After grouping, see aggregate, apply, and transform functions. Here are
some other brief notes about usage. When grouping by multiple groups, the
result index will be a MultiIndex (hierarchical) by default.

Iteration produces (key, group) tuples, i.e. chunking the data by group. So
you can write code like:

::

grouped = obj.groupby(keys, axis=axis)
for key, group in grouped:
# do something with the data

Function calls on GroupBy, if not specially implemented, "dispatch" to the
grouped data. So if you group a DataFrame and wish to invoke the std()
method on each group, you can simply do:

::

df.groupby(mapper).std()

rather than

::

df.groupby(mapper).aggregate(np.std)

You can pass arguments to these "wrapped" functions, too.

See the online documentation for full exposition on these topics and much
more

Returns
-------
**Attributes**
groups : dict
{group name -> group labels}
len(grouped) : int
Number of groups
"""
    _apply_whitelist = _common_apply_whitelist
    _internal_names = ['_cache']
    _internal_names_set = set(_internal_names)
    _group_selection = None

    def __init__(self, obj, keys=None, axis=0, level=None,
                 grouper=None, exclusions=None, selection=None, as_index=True,
                 sort=True, group_keys=True, squeeze=False):
        self._selection = selection

        if isinstance(obj, NDFrame):
            obj._consolidate_inplace()

        self.level = level

        if not as_index:
            if not isinstance(obj, DataFrame):
                raise TypeError('as_index=False only valid with DataFrame')
            if axis != 0:
                raise ValueError('as_index=False only valid for axis=0')

        self.as_index = as_index
        self.keys = keys
        self.sort = sort
        self.group_keys = group_keys
        self.squeeze = squeeze

        if grouper is None:
            grouper, exclusions, obj = _get_grouper(obj, keys, axis=axis,
                                                    level=level, sort=sort)

        self.obj = obj
        self.axis = obj._get_axis_number(axis)
        self.grouper = grouper
        self.exclusions = set(exclusions) if exclusions else set()

    def __len__(self):
        return len(self.indices)

    def __unicode__(self):
        # TODO: Better unicode/repr for GroupBy object
        return object.__repr__(self)

    @property
    def groups(self):
        """ dict {group name -> group labels} """
        return self.grouper.groups

    @property
    def ngroups(self):
        return self.grouper.ngroups

    @property
    def indices(self):
        """ dict {group name -> group indices} """
        return self.grouper.indices

    def _get_index(self, name):
        """ safe get index, translate keys for datelike to underlying repr """

        def convert(key, s):
            # possibly convert to they actual key types
            # in the indices, could be a Timestamp or a np.datetime64

            if isinstance(s, (Timestamp,datetime.datetime)):
                return Timestamp(key)
            elif isinstance(s, np.datetime64):
                return Timestamp(key).asm8
            return key

        sample = next(iter(self.indices))
        if isinstance(sample, tuple):
            if not isinstance(name, tuple):
                raise ValueError("must supply a tuple to get_group with multiple grouping keys")
            if not len(name) == len(sample):
                raise ValueError("must supply a a same-length tuple to get_group with multiple grouping keys")

            name = tuple([ convert(n, k) for n, k in zip(name,sample) ])

        else:

            name = convert(name, sample)

        return self.indices[name]

    @property
    def name(self):
        if self._selection is None:
            return None # 'result'
        else:
            return self._selection

    @property
    def _selection_list(self):
        if not isinstance(self._selection, (list, tuple, Series, np.ndarray)):
            return [self._selection]
        return self._selection

    @cache_readonly
    def _selected_obj(self):

        if self._selection is None or isinstance(self.obj, Series):
            if self._group_selection is not None:
                return self.obj[self._group_selection]
            return self.obj
        else:
            return self.obj[self._selection]

    def _set_selection_from_grouper(self):
        """ we may need create a selection if we have non-level groupers """
        grp = self.grouper
        if self.as_index and getattr(grp,'groupings',None) is not None and self.obj.ndim > 1:
            ax = self.obj._info_axis
            groupers = [ g.name for g in grp.groupings if g.level is None and g.name is not None and g.name in ax ]
            if len(groupers):
                self._group_selection = (ax-Index(groupers)).tolist()

    def _local_dir(self):
        return sorted(set(self.obj._local_dir() + list(self._apply_whitelist)))

    def __getattr__(self, attr):
        if attr in self._internal_names_set:
            return object.__getattribute__(self, attr)
        if attr in self.obj:
            return self[attr]
        if hasattr(self.obj, attr):
            return self._make_wrapper(attr)

        raise AttributeError("%r object has no attribute %r" %
                             (type(self).__name__, attr))

    def __getitem__(self, key):
        raise NotImplementedError('Not implemented: %s' % key)

    def _make_wrapper(self, name):
        if name not in self._apply_whitelist:
            is_callable = callable(getattr(self._selected_obj, name, None))
            kind = ' callable ' if is_callable else ' '
            msg = ("Cannot access{0}attribute {1!r} of {2!r} objects, try "
                   "using the 'apply' method".format(kind, name,
                                                     type(self).__name__))
            raise AttributeError(msg)

        # need to setup the selection
        # as are not passed directly but in the grouper
        self._set_selection_from_grouper()

        f = getattr(self._selected_obj, name)
        if not isinstance(f, types.MethodType):
            return self.apply(lambda self: getattr(self, name))

        f = getattr(type(self._selected_obj), name)

        def wrapper(*args, **kwargs):
            # a little trickery for aggregation functions that need an axis
            # argument
            kwargs_with_axis = kwargs.copy()
            if 'axis' not in kwargs_with_axis:
                kwargs_with_axis['axis'] = self.axis

            def curried_with_axis(x):
                return f(x, *args, **kwargs_with_axis)

            def curried(x):
                return f(x, *args, **kwargs)

            # preserve the name so we can detect it when calling plot methods,
            # to avoid duplicates
            curried.__name__ = curried_with_axis.__name__ = name

            # special case otherwise extra plots are created when catching the
            # exception below
            if name in _plotting_methods:
                return self.apply(curried)

            try:
                return self.apply(curried_with_axis)
            except Exception:
                try:
                    return self.apply(curried)
                except Exception:

                    # related to : GH3688
                    # try item-by-item
                    # this can be called recursively, so need to raise ValueError if
                    # we don't have this method to indicated to aggregate to
                    # mark this column as an error
                    try:
                        return self._aggregate_item_by_item(name, *args, **kwargs)
                    except (AttributeError):
                        raise ValueError

        return wrapper

    def get_group(self, name, obj=None):
        """
Constructs NDFrame from group with provided name

Parameters
----------
name : object
the name of the group to get as a DataFrame
obj : NDFrame, default None
the NDFrame to take the DataFrame out of. If
it is None, the object groupby was called on will
be used

Returns
-------
group : type of obj
"""
        if obj is None:
            obj = self._selected_obj

        inds = self._get_index(name)
        return obj.take(inds, axis=self.axis, convert=False)

    def __iter__(self):
        """
Groupby iterator

Returns
-------
Generator yielding sequence of (name, subsetted object)
for each group
"""
        return self.grouper.get_iterator(self.obj, axis=self.axis)

    def apply(self, func, *args, **kwargs):
        """
Apply function and combine results together in an intelligent way. The
split-apply-combine combination rules attempt to be as common sense
based as possible. For example:

case 1:
group DataFrame
apply aggregation function (f(chunk) -> Series)
yield DataFrame, with group axis having group labels

case 2:
group DataFrame
apply transform function ((f(chunk) -> DataFrame with same indexes)
yield DataFrame with resulting chunks glued together

case 3:
group Series
apply function with f(chunk) -> DataFrame
yield DataFrame with result of chunks glued together

Parameters
----------
func : function

Notes
-----
See online documentation for full exposition on how to use apply.

In the current implementation apply calls func twice on the
first group to decide whether it can take a fast or slow code
path. This can lead to unexpected behavior if func has
side-effects, as they will take effect twice for the first
group.


See also
--------
aggregate, transform

Returns
-------
applied : type depending on grouped object and function
"""
        func = _intercept_function(func)

        @wraps(func)
        def f(g):
            return func(g, *args, **kwargs)

        return self._python_apply_general(f)

    def _python_apply_general(self, f):
        keys, values, mutated = self.grouper.apply(f, self._selected_obj,
                                                   self.axis)

        return self._wrap_applied_output(keys, values,
                                         not_indexed_same=mutated)

    def aggregate(self, func, *args, **kwargs):
        raise NotImplementedError

    @Appender(_agg_doc)
    def agg(self, func, *args, **kwargs):
        return self.aggregate(func, *args, **kwargs)

    def _iterate_slices(self):
        yield self.name, self._selected_obj

    def transform(self, func, *args, **kwargs):
        raise NotImplementedError

    def mean(self):
        """
Compute mean of groups, excluding missing values

For multiple groupings, the result index will be a MultiIndex
"""
        try:
            return self._cython_agg_general('mean')
        except GroupByError:
            raise
        except Exception: # pragma: no cover
            self._set_selection_from_grouper()
            f = lambda x: x.mean(axis=self.axis)
            return self._python_agg_general(f)

    def median(self):
        """
Compute median of groups, excluding missing values

For multiple groupings, the result index will be a MultiIndex
"""
        try:
            return self._cython_agg_general('median')
        except GroupByError:
            raise
        except Exception: # pragma: no cover

            self._set_selection_from_grouper()
            def f(x):
                if isinstance(x, np.ndarray):
                    x = Series(x)
                return x.median(axis=self.axis)
            return self._python_agg_general(f)

    def std(self, ddof=1):
        """
Compute standard deviation of groups, excluding missing values

For multiple groupings, the result index will be a MultiIndex
"""
        # todo, implement at cython level?
        return np.sqrt(self.var(ddof=ddof))

    def var(self, ddof=1):
        """
Compute variance of groups, excluding missing values

For multiple groupings, the result index will be a MultiIndex
"""
        if ddof == 1:
            return self._cython_agg_general('var')
        else:
            self._set_selection_from_grouper()
            f = lambda x: x.var(ddof=ddof)
            return self._python_agg_general(f)

    def sem(self, ddof=1):
        """
Compute standard error of the mean of groups, excluding missing values

For multiple groupings, the result index will be a MultiIndex
"""
        return self.std(ddof=ddof)/np.sqrt(self.count())

    def size(self):
        """
Compute group sizes

"""
        return self.grouper.size()

    sum = _groupby_function('sum', 'add', np.sum)
    prod = _groupby_function('prod', 'prod', np.prod)
    min = _groupby_function('min', 'min', np.min, numeric_only=False)
    max = _groupby_function('max', 'max', np.max, numeric_only=False)
    first = _groupby_function('first', 'first', _first_compat,
                              numeric_only=False, _convert=True)
    last = _groupby_function('last', 'last', _last_compat, numeric_only=False,
                             _convert=True)
    _count = _groupby_function('_count', 'count', _count_compat,
                               numeric_only=False)

    def count(self, axis=0):
        return self._count().astype('int64')

    def ohlc(self):
        """
Compute sum of values, excluding missing values
For multiple groupings, the result index will be a MultiIndex
"""
        return self._apply_to_column_groupbys(
            lambda x: x._cython_agg_general('ohlc'))

    def nth(self, n, dropna=None):
        """
Take the nth row from each group.

If dropna, will not show nth non-null row, dropna is either
Truthy (if a Series) or 'all', 'any' (if a DataFrame); this is equivalent
to calling dropna(how=dropna) before the groupby.

Examples
--------
>>> df = DataFrame([[1, np.nan], [1, 4], [5, 6]], columns=['A', 'B'])
>>> g = df.groupby('A')
>>> g.nth(0)
A B
0 1 NaN
2 5 6
>>> g.nth(1)
A B
1 1 4
>>> g.nth(-1)
A B
1 1 4
2 5 6
>>> g.nth(0, dropna='any')
B
A
1 4
5 6
>>> g.nth(1, dropna='any') # NaNs denote group exhausted when using dropna
B
A
1 NaN
5 NaN

"""

        self._set_selection_from_grouper()
        if not dropna: # good choice
            m = self.grouper._max_groupsize
            if n >= m or n < -m:
                return self._selected_obj.loc[[]]
            rng = np.zeros(m, dtype=bool)
            if n >= 0:
                rng[n] = True
                is_nth = self._cumcount_array(rng)
            else:
                rng[- n - 1] = True
                is_nth = self._cumcount_array(rng, ascending=False)

            result = self._selected_obj[is_nth]

            # the result index
            if self.as_index:
                ax = self.obj._info_axis
                names = self.grouper.names
                if self.obj.ndim == 1:
                    # this is a pass-thru
                    pass
                elif all([ n in ax for n in names ]):
                    result.index = Index(self.obj[names][is_nth].values.ravel()).set_names(names)
                elif self._group_selection is not None:
                    result.index = self.obj._get_axis(self.axis)[is_nth]

                result = result.sort_index()

            return result

        if (isinstance(self._selected_obj, DataFrame)
           and dropna not in ['any', 'all']):
            # Note: when agg-ing picker doesn't raise this, just returns NaN
            raise ValueError("For a DataFrame groupby, dropna must be "
                             "either None, 'any' or 'all', "
                             "(was passed %s)." % (dropna),)

        # old behaviour, but with all and any support for DataFrames.
        # modified in GH 7559 to have better perf
        max_len = n if n >= 0 else - 1 - n
        dropped = self.obj.dropna(how=dropna, axis=self.axis)

        # get a new grouper for our dropped obj
        if self.keys is None and self.level is None:

            # we don't have the grouper info available (e.g. we have selected out
            # a column that is not in the current object)
            axis = self.grouper.axis
            grouper = axis[axis.isin(dropped.index)]
            keys = self.grouper.names
        else:

            # create a grouper with the original parameters, but on the dropped object
            grouper, _, _ = _get_grouper(dropped, key=self.keys, axis=self.axis,
                                         level=self.level, sort=self.sort)

        sizes = dropped.groupby(grouper).size()
        result = dropped.groupby(grouper).nth(n)
        mask = (sizes<max_len).values

        # set the results which don't meet the criteria
        if len(result) and mask.any():
            result.loc[mask] = np.nan

        # reset/reindex to the original groups
        if len(self.obj) == len(dropped) or len(result) == len(self.grouper.result_index):
            result.index = self.grouper.result_index
        else:
            result = result.reindex(self.grouper.result_index)

        return result

    def cumcount(self, **kwargs):
        """
Number each item in each group from 0 to the length of that group - 1.

Essentially this is equivalent to

>>> self.apply(lambda x: Series(np.arange(len(x)), x.index))

Parameters
----------
ascending : bool, default True
If False, number in reverse, from length of group - 1 to 0.

Example
-------

>>> df = pd.DataFrame([['a'], ['a'], ['a'], ['b'], ['b'], ['a']],
... columns=['A'])
>>> df
A
0 a
1 a
2 a
3 b
4 b
5 a
>>> df.groupby('A').cumcount()
0 0
1 1
2 2
3 0
4 1
5 3
dtype: int64
>>> df.groupby('A').cumcount(ascending=False)
0 3
1 2
2 1
3 1
4 0
5 0
dtype: int64

"""
        self._set_selection_from_grouper()
        ascending = kwargs.pop('ascending', True)

        index = self._selected_obj.index
        cumcounts = self._cumcount_array(ascending=ascending)
        return Series(cumcounts, index)

    def head(self, n=5):
        """
Returns first n rows of each group.

Essentially equivalent to ``.apply(lambda x: x.head(n))``,
except ignores as_index flag.

Example
-------

>>> df = DataFrame([[1, 2], [1, 4], [5, 6]],
columns=['A', 'B'])
>>> df.groupby('A', as_index=False).head(1)
A B
0 1 2
2 5 6
>>> df.groupby('A').head(1)
A B
0 1 2
2 5 6

"""
        obj = self._selected_obj
        in_head = self._cumcount_array() < n
        head = obj[in_head]
        return head

    def tail(self, n=5):
        """
Returns last n rows of each group

Essentially equivalent to ``.apply(lambda x: x.tail(n))``,
except ignores as_index flag.

Example
-------

>>> df = DataFrame([[1, 2], [1, 4], [5, 6]],
columns=['A', 'B'])
>>> df.groupby('A', as_index=False).tail(1)
A B
0 1 2
2 5 6
>>> df.groupby('A').head(1)
A B
0 1 2
2 5 6

"""
        obj = self._selected_obj
        rng = np.arange(0, -self.grouper._max_groupsize, -1, dtype='int64')
        in_tail = self._cumcount_array(rng, ascending=False) > -n
        tail = obj[in_tail]
        return tail

    def _cumcount_array(self, arr=None, **kwargs):
        """
arr is where cumcount gets it's values from

note: this is currently implementing sort=False (though the default is sort=True)
for groupby in general
"""
        ascending = kwargs.pop('ascending', True)

        if arr is None:
            arr = np.arange(self.grouper._max_groupsize, dtype='int64')

        len_index = len(self._selected_obj.index)
        cumcounts = np.zeros(len_index, dtype=arr.dtype)
        if not len_index:
            return cumcounts

        indices, values = [], []
        for v in self.indices.values():
            indices.append(v)

            if ascending:
                values.append(arr[:len(v)])
            else:
                values.append(arr[len(v)-1::-1])

        indices = np.concatenate(indices)
        values = np.concatenate(values)
        cumcounts[indices] = values

        return cumcounts

    def _index_with_as_index(self, b):
        """
Take boolean mask of index to be returned from apply, if as_index=True

"""
        # TODO perf, it feels like this should already be somewhere...
        from itertools import chain
        original = self._selected_obj.index
        gp = self.grouper
        levels = chain((gp.levels[i][gp.labels[i][b]]
                        for i in range(len(gp.groupings))),
                       (original.get_level_values(i)[b]
                        for i in range(original.nlevels)))
        new = MultiIndex.from_arrays(list(levels))
        new.names = gp.names + original.names
        return new

    def _try_cast(self, result, obj):
        """
try to cast the result to our obj original type,
we may have roundtripped thru object in the mean-time

"""
        if obj.ndim > 1:
            dtype = obj.values.dtype
        else:
            dtype = obj.dtype

        if not np.isscalar(result):
            result = _possibly_downcast_to_dtype(result, dtype)

        return result

    def _cython_agg_general(self, how, numeric_only=True):
        output = {}
        for name, obj in self._iterate_slices():
            is_numeric = is_numeric_dtype(obj.dtype)
            if numeric_only and not is_numeric:
                continue

            try:
                result, names = self.grouper.aggregate(obj.values, how)
            except AssertionError as e:
                raise GroupByError(str(e))
            output[name] = self._try_cast(result, obj)

        if len(output) == 0:
            raise DataError('No numeric types to aggregate')

        return self._wrap_aggregated_output(output, names)

    def _python_agg_general(self, func, *args, **kwargs):
        func = _intercept_function(func)
        f = lambda x: func(x, *args, **kwargs)

        # iterate through "columns" ex exclusions to populate output dict
        output = {}
        for name, obj in self._iterate_slices():
            try:
                result, counts = self.grouper.agg_series(obj, f)
                output[name] = self._try_cast(result, obj)
            except TypeError:
                continue

        if len(output) == 0:
            return self._python_apply_general(f)

        if self.grouper._filter_empty_groups:

            mask = counts.ravel() > 0
            for name, result in compat.iteritems(output):

                # since we are masking, make sure that we have a float object
                values = result
                if is_numeric_dtype(values.dtype):
                    values = com.ensure_float(values)

                output[name] = self._try_cast(values[mask], result)

        return self._wrap_aggregated_output(output)

    def _wrap_applied_output(self, *args, **kwargs):
        raise NotImplementedError

    def _concat_objects(self, keys, values, not_indexed_same=False):
        from pandas.tools.merge import concat

        if not not_indexed_same:
            result = concat(values, axis=self.axis)
            ax = self._selected_obj._get_axis(self.axis)

            if isinstance(result, Series):
                result = result.reindex(ax)
            else:
                result = result.reindex_axis(ax, axis=self.axis)

        elif self.group_keys:

            if self.as_index:

                # possible MI return case
                group_keys = keys
                group_levels = self.grouper.levels
                group_names = self.grouper.names
                result = concat(values, axis=self.axis, keys=group_keys,
                                levels=group_levels, names=group_names)
            else:

                # GH5610, returns a MI, with the first level being a
                # range index
                keys = list(range(len(values)))
                result = concat(values, axis=self.axis, keys=keys)
        else:
            result = concat(values, axis=self.axis)

        return result

    def _apply_filter(self, indices, dropna):
        if len(indices) == 0:
            indices = []
        else:
            indices = np.sort(np.concatenate(indices))
        if dropna:
            filtered = self._selected_obj.take(indices)
        else:
            mask = np.empty(len(self._selected_obj.index), dtype=bool)
            mask.fill(False)
            mask[indices.astype(int)] = True
            # mask fails to broadcast when passed to where; broadcast manually.
            mask = np.tile(mask, list(self._selected_obj.shape[1:]) + [1]).T
            filtered = self._selected_obj.where(mask) # Fill with NaNs.
        return filtered


@Appender(GroupBy.__doc__)
def groupby(obj, by, **kwds):
    if isinstance(obj, Series):
        klass = SeriesGroupBy
    elif isinstance(obj, DataFrame):
        klass = DataFrameGroupBy
    else: # pragma: no cover
        raise TypeError('invalid type: %s' % type(obj))

    return klass(obj, by, **kwds)


def _get_axes(group):
    if isinstance(group, Series):
        return [group.index]
    else:
        return group.axes


def _is_indexed_like(obj, axes):
    if isinstance(obj, Series):
        if len(axes) > 1:
            return False
        return obj.index.equals(axes[0])
    elif isinstance(obj, DataFrame):
        return obj.index.equals(axes[0])

    return False


class BaseGrouper(object):
    """
This is an internal Grouper class, which actually holds the generated groups
"""

    def __init__(self, axis, groupings, sort=True, group_keys=True):
        self.axis = axis
        self.groupings = groupings
        self.sort = sort
        self.group_keys = group_keys
        self.compressed = True

    @property
    def shape(self):
        return tuple(ping.ngroups for ping in self.groupings)

    def __iter__(self):
        return iter(self.indices)

    @property
    def nkeys(self):
        return len(self.groupings)

    def get_iterator(self, data, axis=0):
        """
Groupby iterator

Returns
-------
Generator yielding sequence of (name, subsetted object)
for each group
"""
        splitter = self._get_splitter(data, axis=axis)
        keys = self._get_group_keys()
        for key, (i, group) in zip(keys, splitter):
            yield key, group

    def _get_splitter(self, data, axis=0):
        comp_ids, _, ngroups = self.group_info
        return get_splitter(data, comp_ids, ngroups, axis=axis)

    def _get_group_keys(self):
        if len(self.groupings) == 1:
            return self.levels[0]
        else:
            comp_ids, _, ngroups = self.group_info
            # provide "flattened" iterator for multi-group setting
            mapper = _KeyMapper(comp_ids, ngroups, self.labels, self.levels)
            return [mapper.get_key(i) for i in range(ngroups)]

    def apply(self, f, data, axis=0):
        mutated = False
        splitter = self._get_splitter(data, axis=axis)
        group_keys = self._get_group_keys()

        # oh boy
        if (f.__name__ not in _plotting_methods and
                hasattr(splitter, 'fast_apply') and axis == 0):
            try:
                values, mutated = splitter.fast_apply(f, group_keys)
                return group_keys, values, mutated
            except (lib.InvalidApply):
                # we detect a mutation of some kind
                # so take slow path
                pass
            except (Exception) as e:
                # raise this error to the caller
                pass

        result_values = []
        for key, (i, group) in zip(group_keys, splitter):
            object.__setattr__(group, 'name', key)

            # group might be modified
            group_axes = _get_axes(group)
            res = f(group)
            if not _is_indexed_like(res, group_axes):
                mutated = True
            result_values.append(res)

        return group_keys, result_values, mutated

    @cache_readonly
    def indices(self):
        """ dict {group name -> group indices} """
        if len(self.groupings) == 1:
            return self.groupings[0].indices
        else:
            label_list = [ping.labels for ping in self.groupings]
            keys = [ping.group_index for ping in self.groupings]
            return _get_indices_dict(label_list, keys)

    @property
    def labels(self):
        return [ping.labels for ping in self.groupings]

    @property
    def levels(self):
        return [ping.group_index for ping in self.groupings]

    @property
    def names(self):
        return [ping.name for ping in self.groupings]

    def size(self):
        """
Compute group sizes

"""
        # TODO: better impl
        labels, _, ngroups = self.group_info
        bin_counts = algos.value_counts(labels, sort=False)
        bin_counts = bin_counts.reindex(np.arange(ngroups))
        bin_counts.index = self.result_index
        return bin_counts

    @cache_readonly
    def _max_groupsize(self):
        '''
Compute size of largest group

'''
        # For many items in each group this is much faster than
        # self.size().max(), in worst case marginally slower
        if self.indices:
            return max(len(v) for v in self.indices.values())
        else:
            return 0

    @cache_readonly
    def groups(self):
        """ dict {group name -> group labels} """
        if len(self.groupings) == 1:
            return self.groupings[0].groups
        else:
            to_groupby = lzip(*(ping.grouper for ping in self.groupings))
            to_groupby = Index(to_groupby)
            return self.axis.groupby(to_groupby.values)

    @cache_readonly
    def group_info(self):
        comp_ids, obs_group_ids = self._get_compressed_labels()

        ngroups = len(obs_group_ids)
        comp_ids = com._ensure_int64(comp_ids)
        return comp_ids, obs_group_ids, ngroups


    def _get_compressed_labels(self):
        all_labels = [ping.labels for ping in self.groupings]
        if self._overflow_possible:
            tups = lib.fast_zip(all_labels)
            labs, uniques = algos.factorize(tups)

            if self.sort:
                uniques, labs = _reorder_by_uniques(uniques, labs)

            return labs, uniques
        else:
            if len(all_labels) > 1:
                group_index = get_group_index(all_labels, self.shape)
                comp_ids, obs_group_ids = _compress_group_index(group_index)
            else:
                ping = self.groupings[0]
                comp_ids = ping.labels
                obs_group_ids = np.arange(len(ping.group_index))
                self.compressed = False
                self._filter_empty_groups = False

            return comp_ids, obs_group_ids

    @cache_readonly
    def _overflow_possible(self):
        return _int64_overflow_possible(self.shape)

    @cache_readonly
    def ngroups(self):
        return len(self.result_index)

    @cache_readonly
    def result_index(self):
        recons = self.get_group_levels()
        return MultiIndex.from_arrays(recons, names=self.names)

    def get_group_levels(self):
        obs_ids = self.group_info[1]

        if not self.compressed and len(self.groupings) == 1:
            return [self.groupings[0].group_index]

        if self._overflow_possible:
            recons_labels = [np.array(x) for x in zip(*obs_ids)]
        else:
            recons_labels = decons_group_index(obs_ids, self.shape)

        name_list = []
        for ping, labels in zip(self.groupings, recons_labels):
            labels = com._ensure_platform_int(labels)
            name_list.append(ping.group_index.take(labels))

        return name_list

    #------------------------------------------------------------
    # Aggregation functions

    _cython_functions = {
        'add': 'group_add',
        'prod': 'group_prod',
        'min': 'group_min',
        'max': 'group_max',
        'mean': 'group_mean',
        'median': {
            'name': 'group_median'
        },
        'var': 'group_var',
        'first': {
            'name': 'group_nth',
            'f': lambda func, a, b, c, d: func(a, b, c, d, 1)
        },
        'last': 'group_last',
        'count': 'group_count',
    }

    _cython_arity = {
        'ohlc': 4, # OHLC
    }

    _name_functions = {}

    _filter_empty_groups = True

    def _get_aggregate_function(self, how, values):

        dtype_str = values.dtype.name

        def get_func(fname):
            # find the function, or use the object function, or return a
            # generic
            for dt in [dtype_str, 'object']:
                f = getattr(_algos, "%s_%s" % (fname, dtype_str), None)
                if f is not None:
                    return f
            return getattr(_algos, fname, None)

        ftype = self._cython_functions[how]

        if isinstance(ftype, dict):
            func = afunc = get_func(ftype['name'])

            # a sub-function
            f = ftype.get('f')
            if f is not None:

                def wrapper(*args, **kwargs):
                    return f(afunc, *args, **kwargs)

                # need to curry our sub-function
                func = wrapper

        else:
            func = get_func(ftype)

        if func is None:
            raise NotImplementedError("function is not implemented for this"
                                      "dtype: [how->%s,dtype->%s]" %
                                      (how, dtype_str))
        return func, dtype_str

    def aggregate(self, values, how, axis=0):

        arity = self._cython_arity.get(how, 1)

        vdim = values.ndim
        swapped = False
        if vdim == 1:
            values = values[:, None]
            out_shape = (self.ngroups, arity)
        else:
            if axis > 0:
                swapped = True
                values = values.swapaxes(0, axis)
            if arity > 1:
                raise NotImplementedError
            out_shape = (self.ngroups,) + values.shape[1:]

        if is_numeric_dtype(values.dtype):
            values = com.ensure_float(values)
            is_numeric = True
            out_dtype = 'f%d' % values.dtype.itemsize
        else:
            is_numeric = issubclass(values.dtype.type, (np.datetime64,
                                                        np.timedelta64))
            if is_numeric:
                out_dtype = 'float64'
                values = values.view('int64')
            else:
                out_dtype = 'object'
                values = values.astype(object)

        # will be filled in Cython function
        result = np.empty(out_shape, dtype=out_dtype)

        result.fill(np.nan)
        counts = np.zeros(self.ngroups, dtype=np.int64)

        result = self._aggregate(result, counts, values, how, is_numeric)

        if self._filter_empty_groups:
            if result.ndim == 2:
                try:
                    result = lib.row_bool_subset(
                        result, (counts > 0).view(np.uint8))
                except ValueError:
                    result = lib.row_bool_subset_object(
                        result, (counts > 0).view(np.uint8))
            else:
                result = result[counts > 0]

        if vdim == 1 and arity == 1:
            result = result[:, 0]

        if how in self._name_functions:
            # TODO
            names = self._name_functions[how]()
        else:
            names = None

        if swapped:
            result = result.swapaxes(0, axis)

        return result, names

    def _aggregate(self, result, counts, values, how, is_numeric):
        agg_func, dtype = self._get_aggregate_function(how, values)

        comp_ids, _, ngroups = self.group_info
        if values.ndim > 3:
            # punting for now
            raise NotImplementedError
        elif values.ndim > 2:
            for i, chunk in enumerate(values.transpose(2, 0, 1)):

                chunk = chunk.squeeze()
                agg_func(result[:, :, i], counts, chunk, comp_ids)
        else:
            agg_func(result, counts, values, comp_ids)

        return result

    def agg_series(self, obj, func):
        try:
            return self._aggregate_series_fast(obj, func)
        except Exception:
            return self._aggregate_series_pure_python(obj, func)

    def _aggregate_series_fast(self, obj, func):
        func = _intercept_function(func)

        if obj.index._has_complex_internals:
            raise TypeError('Incompatible index for Cython grouper')

        group_index, _, ngroups = self.group_info

        # avoids object / Series creation overhead
        dummy = obj._get_values(slice(None, 0)).to_dense()
        indexer = _algos.groupsort_indexer(group_index, ngroups)[0]
        obj = obj.take(indexer, convert=False)
        group_index = com.take_nd(group_index, indexer, allow_fill=False)
        grouper = lib.SeriesGrouper(obj, func, group_index, ngroups,
                                    dummy)
        result, counts = grouper.get_result()
        return result, counts

    def _aggregate_series_pure_python(self, obj, func):

        group_index, _, ngroups = self.group_info

        counts = np.zeros(ngroups, dtype=int)
        result = None

        splitter = get_splitter(obj, group_index, ngroups, axis=self.axis)

        for label, group in splitter:
            res = func(group)
            if result is None:
                if (isinstance(res, (Series, np.ndarray)) or
                        isinstance(res, list)):
                    raise ValueError('Function does not reduce')
                result = np.empty(ngroups, dtype='O')

            counts[label] = group.shape[0]
            result[label] = res

        result = lib.maybe_convert_objects(result, try_float=0)
        return result, counts


def generate_bins_generic(values, binner, closed):
    """
Generate bin edge offsets and bin labels for one array using another array
which has bin edge values. Both arrays must be sorted.

Parameters
----------
values : array of values
binner : a comparable array of values representing bins into which to bin
the first array. Note, 'values' end-points must fall within 'binner'
end-points.
closed : which end of bin is closed; left (default), right

Returns
-------
bins : array of offsets (into 'values' argument) of bins.
Zero and last edge are excluded in result, so for instance the first
bin is values[0:bin[0]] and the last is values[bin[-1]:]
"""
    lenidx = len(values)
    lenbin = len(binner)

    if lenidx <= 0 or lenbin <= 0:
        raise ValueError("Invalid length for values or for binner")

    # check binner fits data
    if values[0] < binner[0]:
        raise ValueError("Values falls before first bin")

    if values[lenidx - 1] > binner[lenbin - 1]:
        raise ValueError("Values falls after last bin")

    bins = np.empty(lenbin - 1, dtype=np.int64)

    j = 0 # index into values
    bc = 0 # bin count

    # linear scan, presume nothing about values/binner except that it fits ok
    for i in range(0, lenbin - 1):
        r_bin = binner[i + 1]

        # count values in current bin, advance to next bin
        while j < lenidx and (values[j] < r_bin or
                              (closed == 'right' and values[j] == r_bin)):
            j += 1

        bins[bc] = j
        bc += 1

    return bins

class BinGrouper(BaseGrouper):

    def __init__(self, bins, binlabels, filter_empty=False):
        self.bins = com._ensure_int64(bins)
        self.binlabels = _ensure_index(binlabels)
        self._filter_empty_groups = filter_empty

    @cache_readonly
    def groups(self):
        """ dict {group name -> group labels} """

        # this is mainly for compat
        # GH 3881
        result = {}
        for key, value in zip(self.binlabels, self.bins):
            if key is not tslib.NaT:
                result[key] = value
        return result

    @property
    def nkeys(self):
        return 1

    def get_iterator(self, data, axis=0):
        """
Groupby iterator

Returns
-------
Generator yielding sequence of (name, subsetted object)
for each group
"""
        if isinstance(data, NDFrame):
            slicer = lambda start,edge: data._slice(slice(start,edge),axis=axis)
            length = len(data.axes[axis])
        else:
            slicer = lambda start,edge: data[slice(start,edge)]
            length = len(data)

        start = 0
        for edge, label in zip(self.bins, self.binlabels):
            if label is not tslib.NaT:
                yield label, slicer(start,edge)
            start = edge

        if start < length:
            yield self.binlabels[-1], slicer(start,None)

    def apply(self, f, data, axis=0):
        result_keys = []
        result_values = []
        mutated = False
        for key, group in self.get_iterator(data, axis=axis):
            object.__setattr__(group, 'name', key)

            # group might be modified
            group_axes = _get_axes(group)
            res = f(group)

            if not _is_indexed_like(res, group_axes):
                mutated = True

            result_keys.append(key)
            result_values.append(res)

        return result_keys, result_values, mutated

    @cache_readonly
    def indices(self):
        indices = collections.defaultdict(list)

        i = 0
        for label, bin in zip(self.binlabels, self.bins):
            if i < bin:
                if label is not tslib.NaT:
                    indices[label] = list(range(i, bin))
                i = bin
        return indices

    @cache_readonly
    def ngroups(self):
        return len(self.binlabels)

    @cache_readonly
    def result_index(self):
        mask = self.binlabels.asi8 == tslib.iNaT
        return self.binlabels[~mask]

    @property
    def levels(self):
        return [self.binlabels]

    @property
    def names(self):
        return [self.binlabels.name]

    def size(self):
        """
Compute group sizes

"""
        base = Series(np.zeros(len(self.result_index), dtype=np.int64),
                      index=self.result_index)
        indices = self.indices
        for k, v in compat.iteritems(indices):
            indices[k] = len(v)
        bin_counts = Series(indices, dtype=np.int64)
        result = base.add(bin_counts, fill_value=0)
        # addition with fill_value changes dtype to float64
        result = result.astype(np.int64)
        return result

    #----------------------------------------------------------------------
    # cython aggregation

    _cython_functions = {
        'add': 'group_add_bin',
        'prod': 'group_prod_bin',
        'mean': 'group_mean_bin',
        'min': 'group_min_bin',
        'max': 'group_max_bin',
        'var': 'group_var_bin',
        'ohlc': 'group_ohlc',
        'first': {
            'name': 'group_nth_bin',
            'f': lambda func, a, b, c, d: func(a, b, c, d, 1)
        },
        'last': 'group_last_bin',
        'count': 'group_count_bin',
    }

    _name_functions = {
        'ohlc': lambda *args: ['open', 'high', 'low', 'close']
    }

    _filter_empty_groups = True

    def _aggregate(self, result, counts, values, how, is_numeric=True):

        agg_func, dtype = self._get_aggregate_function(how, values)

        if values.ndim > 3:
            # punting for now
            raise NotImplementedError
        elif values.ndim > 2:
            for i, chunk in enumerate(values.transpose(2, 0, 1)):
                agg_func(result[:, :, i], counts, chunk, self.bins)
        else:
            agg_func(result, counts, values, self.bins)

        return result

    def agg_series(self, obj, func):
        dummy = obj[:0]
        grouper = lib.SeriesBinGrouper(obj, func, self.bins, dummy)
        return grouper.get_result()


class Grouping(object):

    """
Holds the grouping information for a single key

Parameters
----------
index : Index
grouper :
obj :
name :
level :

Returns
-------
**Attributes**:
* indices : dict of {group -> index_list}
* labels : ndarray, group labels
* ids : mapping of label -> group
* counts : array of group counts
* group_index : unique groups
* groups : dict of {group -> label_list}
"""

    def __init__(self, index, grouper=None, obj=None, name=None, level=None,
                 sort=True):

        self.name = name
        self.level = level
        self.grouper = _convert_grouper(index, grouper)
        self.index = index
        self.sort = sort
        self.obj = obj

        # right place for this?
        if isinstance(grouper, (Series, Index)) and name is None:
            self.name = grouper.name

        if isinstance(grouper, MultiIndex):
            self.grouper = grouper.values

        # pre-computed
        self._was_factor = False
        self._should_compress = True

        # we have a single grouper which may be a myriad of things, some of which are
        # dependent on the passing in level
        #

        if level is not None:
            if not isinstance(level, int):
                if level not in index.names:
                    raise AssertionError('Level %s not in index' % str(level))
                level = index.names.index(level)

            inds = index.labels[level]
            level_index = index.levels[level]

            if self.name is None:
                self.name = index.names[level]

            # XXX complete hack

            if grouper is not None:
                level_values = index.levels[level].take(inds)
                self.grouper = level_values.map(self.grouper)
            else:
                self._was_factor = True

                # all levels may not be observed
                labels, uniques = algos.factorize(inds, sort=True)

                if len(uniques) > 0 and uniques[0] == -1:
                    # handle NAs
                    mask = inds != -1
                    ok_labels, uniques = algos.factorize(inds[mask], sort=True)

                    labels = np.empty(len(inds), dtype=inds.dtype)
                    labels[mask] = ok_labels
                    labels[~mask] = -1

                if len(uniques) < len(level_index):
                    level_index = level_index.take(uniques)

                self._labels = labels
                self._group_index = level_index
                self.grouper = level_index.take(labels)
        else:
            if isinstance(self.grouper, (list, tuple)):
                self.grouper = com._asarray_tuplesafe(self.grouper)

            # a passed Categorical
            elif isinstance(self.grouper, Categorical):

                factor = self.grouper
                self._was_factor = True

                # Is there any way to avoid this?
                self.grouper = np.asarray(factor)

                self._labels = factor.labels
                self._group_index = factor.levels
                if self.name is None:
                    self.name = factor.name

            # a passed Grouper like
            elif isinstance(self.grouper, Grouper):

                # get the new grouper
                grouper = self.grouper._get_binner_for_grouping(self.obj)
                self.obj = self.grouper.obj
                self.grouper = grouper
                if self.name is None:
                    self.name = grouper.name

            # no level passed
            if not isinstance(self.grouper, (Series, np.ndarray)):
                self.grouper = self.index.map(self.grouper)
                if not (hasattr(self.grouper, "__len__") and
                        len(self.grouper) == len(self.index)):
                    errmsg = ('Grouper result violates len(labels) == '
                              'len(data)\nresult: %s' %
                              com.pprint_thing(self.grouper))
                    self.grouper = None # Try for sanity
                    raise AssertionError(errmsg)

        # if we have a date/time-like grouper, make sure that we have Timestamps like
        if getattr(self.grouper,'dtype',None) is not None:
            if is_datetime64_dtype(self.grouper):
                from pandas import to_datetime
                self.grouper = to_datetime(self.grouper)
            elif is_timedelta64_dtype(self.grouper):
                from pandas import to_timedelta
                self.grouper = to_timedelta(self.grouper)

    def __repr__(self):
        return 'Grouping(%s)' % self.name

    def __iter__(self):
        return iter(self.indices)

    _labels = None
    _group_index = None

    @property
    def ngroups(self):
        return len(self.group_index)

    @cache_readonly
    def indices(self):
        return _groupby_indices(self.grouper)

    @property
    def labels(self):
        if self._labels is None:
            self._make_labels()
        return self._labels

    @property
    def group_index(self):
        if self._group_index is None:
            self._make_labels()
        return self._group_index

    def _make_labels(self):
        if self._was_factor: # pragma: no cover
            raise Exception('Should not call this method grouping by level')
        else:
            labels, uniques = algos.factorize(self.grouper, sort=self.sort)
            uniques = Index(uniques, name=self.name)
            self._labels = labels
            self._group_index = uniques

    _groups = None

    @property
    def groups(self):
        if self._groups is None:
            self._groups = self.index.groupby(self.grouper)
        return self._groups

def _get_grouper(obj, key=None, axis=0, level=None, sort=True):
    """
create and return a BaseGrouper, which is an internal
mapping of how to create the grouper indexers.
This may be composed of multiple Grouping objects, indicating
multiple groupers

Groupers are ultimately index mappings. They can originate as:
index mappings, keys to columns, functions, or Groupers

Groupers enable local references to axis,level,sort, while
the passed in axis, level, and sort are 'global'.

This routine tries to figure of what the passing in references
are and then creates a Grouping for each one, combined into
a BaseGrouper.

"""

    group_axis = obj._get_axis(axis)

    # validate thatthe passed level is compatible with the passed
    # axis of the object
    if level is not None:
        if not isinstance(group_axis, MultiIndex):
            if isinstance(level, compat.string_types):
                if obj.index.name != level:
                    raise ValueError('level name %s is not the name of the '
                                     'index' % level)
            elif level > 0:
                raise ValueError('level > 0 only valid with MultiIndex')

            level = None
            key = group_axis

    # a passed in Grouper, directly convert
    if isinstance(key, Grouper):
        binner, grouper, obj = key._get_grouper(obj)
        if key.key is None:
            return grouper, [], obj
        else:
            return grouper, set([key.key]), obj

    # already have a BaseGrouper, just return it
    elif isinstance(key, BaseGrouper):
        return key, [], obj

    if not isinstance(key, (tuple, list)):
        keys = [key]
    else:
        keys = key

    # what are we after, exactly?
    match_axis_length = len(keys) == len(group_axis)
    any_callable = any(callable(g) or isinstance(g, dict) for g in keys)
    any_arraylike = any(isinstance(g, (list, tuple, Series, np.ndarray))
                        for g in keys)

    try:
        if isinstance(obj, DataFrame):
            all_in_columns = all(g in obj.columns for g in keys)
        else:
            all_in_columns = False
    except Exception:
        all_in_columns = False

    if (not any_callable and not all_in_columns
        and not any_arraylike and match_axis_length
            and level is None):
        keys = [com._asarray_tuplesafe(keys)]

    if isinstance(level, (tuple, list)):
        if key is None:
            keys = [None] * len(level)
        levels = level
    else:
        levels = [level] * len(keys)

    groupings = []
    exclusions = []
    for i, (gpr, level) in enumerate(zip(keys, levels)):
        name = None
        try:
            obj._data.items.get_loc(gpr)
            in_axis = True
        except Exception:
            in_axis = False

        if _is_label_like(gpr) or in_axis:
            exclusions.append(gpr)
            name = gpr
            gpr = obj[gpr]

        if isinstance(gpr, Categorical) and len(gpr) != len(obj):
            errmsg = "Categorical grouper must have len(grouper) == len(data)"
            raise AssertionError(errmsg)

        ping = Grouping(group_axis, gpr, obj=obj, name=name, level=level, sort=sort)
        groupings.append(ping)

    if len(groupings) == 0:
        raise ValueError('No group keys passed!')

    # create the internals grouper
    grouper = BaseGrouper(group_axis, groupings, sort=sort)

    return grouper, exclusions, obj


def _is_label_like(val):
    return isinstance(val, compat.string_types) or np.isscalar(val)


def _convert_grouper(axis, grouper):
    if isinstance(grouper, dict):
        return grouper.get
    elif isinstance(grouper, Series):
        if grouper.index.equals(axis):
            return grouper.values
        else:
            return grouper.reindex(axis).values
    elif isinstance(grouper, (list, Series, np.ndarray)):
        if len(grouper) != len(axis):
            raise AssertionError('Grouper and axis must be same length')
        return grouper
    else:
        return grouper


class SeriesGroupBy(GroupBy):
    _apply_whitelist = _series_apply_whitelist

    def aggregate(self, func_or_funcs, *args, **kwargs):
        """
Apply aggregation function or functions to groups, yielding most likely
Series but in some cases DataFrame depending on the output of the
aggregation function

Parameters
----------
func_or_funcs : function or list / dict of functions
List/dict of functions will produce DataFrame with column names
determined by the function names themselves (list) or the keys in
the dict

Notes
-----
agg is an alias for aggregate. Use it.

Examples
--------
>>> series
bar 1.0
baz 2.0
qot 3.0
qux 4.0

>>> mapper = lambda x: x[0] # first letter
>>> grouped = series.groupby(mapper)

>>> grouped.aggregate(np.sum)
b 3.0
q 7.0

>>> grouped.aggregate([np.sum, np.mean, np.std])
mean std sum
b 1.5 0.5 3
q 3.5 0.5 7

>>> grouped.agg({'result' : lambda x: x.mean() / x.std(),
... 'total' : np.sum})
result total
b 2.121 3
q 4.95 7

See also
--------
apply, transform

Returns
-------
Series or DataFrame
"""
        if isinstance(func_or_funcs, compat.string_types):
            return getattr(self, func_or_funcs)(*args, **kwargs)

        if hasattr(func_or_funcs, '__iter__'):
            ret = self._aggregate_multiple_funcs(func_or_funcs)
        else:
            cyfunc = _intercept_cython(func_or_funcs)
            if cyfunc and not args and not kwargs:
                return getattr(self, cyfunc)()

            if self.grouper.nkeys > 1:
                return self._python_agg_general(func_or_funcs, *args, **kwargs)

            try:
                return self._python_agg_general(func_or_funcs, *args, **kwargs)
            except Exception:
                result = self._aggregate_named(func_or_funcs, *args, **kwargs)

            index = Index(sorted(result), name=self.grouper.names[0])
            ret = Series(result, index=index)

        if not self.as_index: # pragma: no cover
            print('Warning, ignoring as_index=True')

        return ret

    def _aggregate_multiple_funcs(self, arg):
        if isinstance(arg, dict):
            columns = list(arg.keys())
            arg = list(arg.items())
        elif any(isinstance(x, (tuple, list)) for x in arg):
            arg = [(x, x) if not isinstance(x, (tuple, list)) else x
                   for x in arg]

            # indicated column order
            columns = lzip(*arg)[0]
        else:
            # list of functions / function names
            columns = []
            for f in arg:
                if isinstance(f, compat.string_types):
                    columns.append(f)
                else:
                    columns.append(f.__name__)
            arg = lzip(columns, arg)

        results = {}

        for name, func in arg:
            if name in results:
                raise SpecificationError('Function names must be unique, '
                                         'found multiple named %s' % name)

            results[name] = self.aggregate(func)

        return DataFrame(results, columns=columns)

    def _wrap_aggregated_output(self, output, names=None):
        # sort of a kludge
        output = output[self.name]
        index = self.grouper.result_index

        if names is not None:
            return DataFrame(output, index=index, columns=names)
        else:
            name = self.name
            if name is None:
                name = self._selected_obj.name
            return Series(output, index=index, name=name)

    def _wrap_applied_output(self, keys, values, not_indexed_same=False):
        if len(keys) == 0:
            # GH #6265
            return Series([], name=self.name)

        def _get_index():
            if self.grouper.nkeys > 1:
                index = MultiIndex.from_tuples(keys, names=self.grouper.names)
            else:
                index = Index(keys, name=self.grouper.names[0])
            return index

        if isinstance(values[0], dict):
            # GH #823
            index = _get_index()
            return DataFrame(values, index=index).stack()

        if isinstance(values[0], (Series, dict)):
            return self._concat_objects(keys, values,
                                        not_indexed_same=not_indexed_same)
        elif isinstance(values[0], DataFrame):
            # possible that Series -> DataFrame by applied function
            return self._concat_objects(keys, values,
                                        not_indexed_same=not_indexed_same)
        else:
            # GH #6265
            return Series(values, index=_get_index(), name=self.name)

    def _aggregate_named(self, func, *args, **kwargs):
        result = {}

        for name, group in self:
            group.name = name
            output = func(group, *args, **kwargs)
            if isinstance(output, (Series, np.ndarray)):
                raise Exception('Must produce aggregated value')
            result[name] = self._try_cast(output, group)

        return result

    def transform(self, func, *args, **kwargs):
        """
Call function producing a like-indexed Series on each group and return
a Series with the transformed values

Parameters
----------
func : function
To apply to each group. Should return a Series with the same index

Examples
--------
>>> grouped.transform(lambda x: (x - x.mean()) / x.std())

Returns
-------
transformed : Series
"""
        dtype = self._selected_obj.dtype

        if isinstance(func, compat.string_types):
            wrapper = lambda x: getattr(x, func)(*args, **kwargs)
        else:
            wrapper = lambda x: func(x, *args, **kwargs)

        result = self._selected_obj.values.copy()
        for i, (name, group) in enumerate(self):

            object.__setattr__(group, 'name', name)
            res = wrapper(group)

            if hasattr(res, 'values'):
                res = res.values

            # may need to astype
            try:
                common_type = np.common_type(np.array(res), result)
                if common_type != result.dtype:
                    result = result.astype(common_type)
            except:
                pass

            indexer = self._get_index(name)
            result[indexer] = res

        result = _possibly_downcast_to_dtype(result, dtype)
        return self._selected_obj.__class__(result,
                                            index=self._selected_obj.index,
                                            name=self._selected_obj.name)

    def filter(self, func, dropna=True, *args, **kwargs):
        """
Return a copy of a Series excluding elements from groups that
do not satisfy the boolean criterion specified by func.

Parameters
----------
func : function
To apply to each group. Should return True or False.
dropna : Drop groups that do not pass the filter. True by default;
if False, groups that evaluate False are filled with NaNs.

Example
-------
>>> grouped.filter(lambda x: x.mean() > 0)

Returns
-------
filtered : Series
"""
        if isinstance(func, compat.string_types):
            wrapper = lambda x: getattr(x, func)(*args, **kwargs)
        else:
            wrapper = lambda x: func(x, *args, **kwargs)

        # Interpret np.nan as False.
        def true_and_notnull(x, *args, **kwargs):
            b = wrapper(x, *args, **kwargs)
            return b and notnull(b)

        try:
            indices = [self._get_index(name) if true_and_notnull(group) else []
                       for name, group in self]
        except ValueError:
            raise TypeError("the filter must return a boolean result")
        except TypeError:
            raise TypeError("the filter must return a boolean result")

        filtered = self._apply_filter(indices, dropna)
        return filtered

    def _apply_to_column_groupbys(self, func):
        """ return a pass thru """
        return func(self)

class NDFrameGroupBy(GroupBy):

    def _iterate_slices(self):
        if self.axis == 0:
            # kludge
            if self._selection is None:
                slice_axis = self.obj.columns
            else:
                slice_axis = self._selection_list
            slicer = lambda x: self.obj[x]
        else:
            slice_axis = self.obj.index
            slicer = self.obj.xs

        for val in slice_axis:
            if val in self.exclusions:
                continue
            yield val, slicer(val)

    def _cython_agg_general(self, how, numeric_only=True):
        new_items, new_blocks = self._cython_agg_blocks(how, numeric_only=numeric_only)
        return self._wrap_agged_blocks(new_items, new_blocks)

    def _wrap_agged_blocks(self, items, blocks):
        obj = self._obj_with_exclusions

        new_axes = list(obj._data.axes)

        # more kludge
        if self.axis == 0:
            new_axes[0], new_axes[1] = new_axes[1], self.grouper.result_index
        else:
            new_axes[self.axis] = self.grouper.result_index

        # Make sure block manager integrity check passes.
        assert new_axes[0].equals(items)
        new_axes[0] = items

        mgr = BlockManager(blocks, new_axes)

        new_obj = type(obj)(mgr)

        return self._post_process_cython_aggregate(new_obj)

    _block_agg_axis = 0

    def _cython_agg_blocks(self, how, numeric_only=True):
        data, agg_axis = self._get_data_to_aggregate()

        new_blocks = []

        if numeric_only:
            data = data.get_numeric_data(copy=False)

        for block in data.blocks:

            values = block._try_operate(block.values)

            if block.is_numeric:
                values = com.ensure_float(values)

            result, _ = self.grouper.aggregate(values, how, axis=agg_axis)

            # see if we can cast the block back to the original dtype
            result = block._try_coerce_and_cast_result(result)

            newb = make_block(result, placement=block.mgr_locs)
            new_blocks.append(newb)

        if len(new_blocks) == 0:
            raise DataError('No numeric types to aggregate')

        return data.items, new_blocks

    def _get_data_to_aggregate(self):
        obj = self._obj_with_exclusions
        if self.axis == 0:
            return obj.swapaxes(0, 1)._data, 1
        else:
            return obj._data, self.axis

    def _post_process_cython_aggregate(self, obj):
        # undoing kludge from below
        if self.axis == 0:
            obj = obj.swapaxes(0, 1)
        return obj

    @cache_readonly
    def _obj_with_exclusions(self):
        if self._selection is not None:
            return self.obj.reindex(columns=self._selection_list)

        if len(self.exclusions) > 0:
            return self.obj.drop(self.exclusions, axis=1)
        else:
            return self.obj

    @Appender(_agg_doc)
    def aggregate(self, arg, *args, **kwargs):
        if isinstance(arg, compat.string_types):
            return getattr(self, arg)(*args, **kwargs)

        result = OrderedDict()
        if isinstance(arg, dict):
            if self.axis != 0: # pragma: no cover
                raise ValueError('Can only pass dict with axis=0')

            obj = self._selected_obj

            if any(isinstance(x, (list, tuple, dict)) for x in arg.values()):
                new_arg = OrderedDict()
                for k, v in compat.iteritems(arg):
                    if not isinstance(v, (tuple, list, dict)):
                        new_arg[k] = [v]
                    else:
                        new_arg[k] = v
                arg = new_arg

            keys = []
            if self._selection is not None:
                subset = obj
                if isinstance(subset, DataFrame):
                    raise NotImplementedError

                for fname, agg_how in compat.iteritems(arg):
                    colg = SeriesGroupBy(subset, selection=self._selection,
                                         grouper=self.grouper)
                    result[fname] = colg.aggregate(agg_how)
                    keys.append(fname)
            else:
                for col, agg_how in compat.iteritems(arg):
                    colg = SeriesGroupBy(obj[col], selection=col,
                                         grouper=self.grouper)
                    result[col] = colg.aggregate(agg_how)
                    keys.append(col)

            if isinstance(list(result.values())[0], DataFrame):
                from pandas.tools.merge import concat
                result = concat([result[k] for k in keys], keys=keys, axis=1)
            else:
                result = DataFrame(result)
        elif isinstance(arg, list):
            return self._aggregate_multiple_funcs(arg)
        else:
            cyfunc = _intercept_cython(arg)
            if cyfunc and not args and not kwargs:
                return getattr(self, cyfunc)()

            if self.grouper.nkeys > 1:
                return self._python_agg_general(arg, *args, **kwargs)
            else:

                # try to treat as if we are passing a list
                try:
                    assert not args and not kwargs
                    result = self._aggregate_multiple_funcs([arg])
                    result.columns = Index(result.columns.levels[0],
                                           name=self._selected_obj.columns.name)
                except:
                    result = self._aggregate_generic(arg, *args, **kwargs)

        if not self.as_index:
            if isinstance(result.index, MultiIndex):
                zipped = zip(result.index.levels, result.index.labels,
                             result.index.names)
                for i, (lev, lab, name) in enumerate(zipped):
                    result.insert(i, name,
                                  com.take_nd(lev.values, lab,
                                              allow_fill=False))
                result = result.consolidate()
            else:
                values = result.index.values
                name = self.grouper.groupings[0].name
                result.insert(0, name, values)
            result.index = np.arange(len(result))

        return result.convert_objects()

    def _aggregate_multiple_funcs(self, arg):
        from pandas.tools.merge import concat

        if self.axis != 0:
            raise NotImplementedError

        obj = self._obj_with_exclusions

        results = []
        keys = []
        for col in obj:
            try:
                colg = SeriesGroupBy(obj[col], selection=col,
                                     grouper=self.grouper)
                results.append(colg.aggregate(arg))
                keys.append(col)
            except (TypeError, DataError):
                pass
            except SpecificationError:
                raise
        result = concat(results, keys=keys, axis=1)

        return result

    def _aggregate_generic(self, func, *args, **kwargs):
        if self.grouper.nkeys != 1:
            raise AssertionError('Number of keys must be 1')

        axis = self.axis
        obj = self._obj_with_exclusions

        result = {}
        if axis != obj._info_axis_number:
            try:
                for name, data in self:
                    # for name in self.indices:
                    # data = self.get_group(name, obj=obj)
                    result[name] = self._try_cast(func(data, *args, **kwargs),
                                                  data)
            except Exception:
                return self._aggregate_item_by_item(func, *args, **kwargs)
        else:
            for name in self.indices:
                try:
                    data = self.get_group(name, obj=obj)
                    result[name] = self._try_cast(func(data, *args, **kwargs),
                                                  data)
                except Exception:
                    wrapper = lambda x: func(x, *args, **kwargs)
                    result[name] = data.apply(wrapper, axis=axis)

        return self._wrap_generic_output(result, obj)

    def _wrap_aggregated_output(self, output, names=None):
        raise NotImplementedError

    def _aggregate_item_by_item(self, func, *args, **kwargs):
        # only for axis==0

        obj = self._obj_with_exclusions
        result = {}
        cannot_agg = []
        errors=None
        for item in obj:
            try:
                data = obj[item]
                colg = SeriesGroupBy(data, selection=item,
                                     grouper=self.grouper)
                result[item] = self._try_cast(
                    colg.aggregate(func, *args, **kwargs), data)
            except ValueError:
                cannot_agg.append(item)
                continue
            except TypeError as e:
                cannot_agg.append(item)
                errors=e
                continue

        result_columns = obj.columns
        if cannot_agg:
            result_columns = result_columns.drop(cannot_agg)

            # GH6337
            if not len(result_columns) and errors is not None:
                raise errors

        return DataFrame(result, columns=result_columns)

    def _decide_output_index(self, output, labels):
        if len(output) == len(labels):
            output_keys = labels
        else:
            output_keys = sorted(output)
            try:
                output_keys.sort()
            except Exception: # pragma: no cover
                pass

            if isinstance(labels, MultiIndex):
                output_keys = MultiIndex.from_tuples(output_keys,
                                                     names=labels.names)

        return output_keys

    def _wrap_applied_output(self, keys, values, not_indexed_same=False):
        from pandas.core.index import _all_indexes_same

        if len(keys) == 0:
            # XXX
            return DataFrame({})

        key_names = self.grouper.names

        if isinstance(values[0], DataFrame):
            return self._concat_objects(keys, values,
                                        not_indexed_same=not_indexed_same)
        elif hasattr(self.grouper, 'groupings'):
            if len(self.grouper.groupings) > 1:
                key_index = MultiIndex.from_tuples(keys, names=key_names)

            else:
                ping = self.grouper.groupings[0]
                if len(keys) == ping.ngroups:
                    key_index = ping.group_index
                    key_index.name = key_names[0]

                    key_lookup = Index(keys)
                    indexer = key_lookup.get_indexer(key_index)

                    # reorder the values
                    values = [values[i] for i in indexer]
                else:

                    key_index = Index(keys, name=key_names[0])

                # don't use the key indexer
                if not self.as_index:
                    key_index = None

            # make Nones an empty object
            if com._count_not_none(*values) != len(values):
                v = next(v for v in values if v is not None)
                if v is None:
                    return DataFrame()
                elif isinstance(v, NDFrame):
                    values = [
                        x if x is not None else
                        v._constructor(**v._construct_axes_dict())
                        for x in values
                        ]

            v = values[0]

            if isinstance(v, (np.ndarray, Series)):
                if isinstance(v, Series):
                    applied_index = self._selected_obj._get_axis(self.axis)
                    all_indexed_same = _all_indexes_same([
                        x.index for x in values
                    ])
                    singular_series = (len(values) == 1 and
                                       applied_index.nlevels == 1)

                    # GH3596
                    # provide a reduction (Frame -> Series) if groups are
                    # unique
                    if self.squeeze:

                        # assign the name to this series
                        if singular_series:
                            values[0].name = keys[0]

                            # GH2893
                            # we have series in the values array, we want to
                            # produce a series:
                            # if any of the sub-series are not indexed the same
                            # OR we don't have a multi-index and we have only a
                            # single values
                            return self._concat_objects(
                                keys, values, not_indexed_same=not_indexed_same
                            )

                        # still a series
                        # path added as of GH 5545
                        elif all_indexed_same:
                            from pandas.tools.merge import concat
                            return concat(values)

                    if not all_indexed_same:
                        return self._concat_objects(
                            keys, values, not_indexed_same=not_indexed_same
                        )

                try:
                    if self.axis == 0:
                        # GH6124 if the list of Series have a consistent name,
                        # then propagate that name to the result.
                        index = v.index.copy()
                        if index.name is None:
                            # Only propagate the series name to the result
                            # if all series have a consistent name. If the
                            # series do not have a consistent name, do
                            # nothing.
                            names = set(v.name for v in values)
                            if len(names) == 1:
                                index.name = list(names)[0]

                        # normally use vstack as its faster than concat
                        # and if we have mi-columns
                        if not _np_version_under1p7 or isinstance(v.index,MultiIndex) or key_index is None:
                            stacked_values = np.vstack([np.asarray(x) for x in values])
                            result = DataFrame(stacked_values,index=key_index,columns=index)
                        else:
                            # GH5788 instead of stacking; concat gets the dtypes correct
                            from pandas.tools.merge import concat
                            result = concat(values,keys=key_index,names=key_index.names,
                                            axis=self.axis).unstack()
                            result.columns = index
                    else:
                        stacked_values = np.vstack([np.asarray(x) for x in values])
                        result = DataFrame(stacked_values.T,index=v.index,columns=key_index)

                except (ValueError, AttributeError):
                    # GH1738: values is list of arrays of unequal lengths fall
                    # through to the outer else caluse
                    return Series(values, index=key_index)

                # if we have date/time like in the original, then coerce dates
                # as we are stacking can easily have object dtypes here
                if (self._selected_obj.ndim == 2
                       and self._selected_obj.dtypes.isin(_DATELIKE_DTYPES).any()):
                    cd = 'coerce'
                else:
                    cd = True
                return result.convert_objects(convert_dates=cd)

            else:
                # only coerce dates if we find at least 1 datetime
                cd = 'coerce' if any([ isinstance(v,Timestamp) for v in values ]) else False
                return Series(values, index=key_index).convert_objects(convert_dates=cd)

        else:
            # Handle cases like BinGrouper
            return self._concat_objects(keys, values,
                                        not_indexed_same=not_indexed_same)

    def _transform_general(self, func, *args, **kwargs):
        from pandas.tools.merge import concat

        applied = []

        obj = self._obj_with_exclusions
        gen = self.grouper.get_iterator(obj, axis=self.axis)
        fast_path, slow_path = self._define_paths(func, *args, **kwargs)

        path = None
        for name, group in gen:
            object.__setattr__(group, 'name', name)

            if path is None:
                # Try slow path and fast path.
                try:
                    path, res = self._choose_path(fast_path, slow_path, group)
                except TypeError:
                    return self._transform_item_by_item(obj, fast_path)
                except Exception: # pragma: no cover
                    res = fast_path(group)
                    path = fast_path
            else:
                res = path(group)

            # broadcasting
            if isinstance(res, Series):
                if res.index.is_(obj.index):
                    group.T.values[:] = res
                else:
                    group.values[:] = res

                applied.append(group)
            else:
                applied.append(res)

        concat_index = obj.columns if self.axis == 0 else obj.index
        concatenated = concat(applied, join_axes=[concat_index],
                              axis=self.axis, verify_integrity=False)
        concatenated.sort_index(inplace=True)
        return concatenated

    def transform(self, func, *args, **kwargs):
        """
Call function producing a like-indexed DataFrame on each group and
return a DataFrame having the same indexes as the original object
filled with the transformed values

Parameters
----------
f : function
Function to apply to each subframe

Notes
-----
Each subframe is endowed the attribute 'name' in case you need to know
which group you are working on.

Examples
--------
>>> grouped = df.groupby(lambda x: mapping[x])
>>> grouped.transform(lambda x: (x - x.mean()) / x.std())
"""

        # try to do a fast transform via merge if possible
        try:
            obj = self._obj_with_exclusions
            if isinstance(func, compat.string_types):
                result = getattr(self, func)(*args, **kwargs)
            else:
                cyfunc = _intercept_cython(func)
                if cyfunc and not args and not kwargs:
                    result = getattr(self, cyfunc)()
                else:
                    return self._transform_general(func, *args, **kwargs)
        except:
            return self._transform_general(func, *args, **kwargs)

        # a reduction transform
        if not isinstance(result, DataFrame):
            return self._transform_general(func, *args, **kwargs)

        # nuiscance columns
        if not result.columns.equals(obj.columns):
            return self._transform_general(func, *args, **kwargs)

        # a grouped that doesn't preserve the index, remap index based on the grouper
        # and broadcast it
        if not isinstance(obj.index,MultiIndex) and type(result.index) != type(obj.index):
            results = obj.values.copy()
            for (name, group), (i, row) in zip(self, result.iterrows()):
                indexer = self._get_index(name)
                results[indexer] = np.tile(row.values,len(indexer)).reshape(len(indexer),-1)
            return DataFrame(results,columns=result.columns,index=obj.index).convert_objects()

        # we can merge the result in
        # GH 7383
        names = result.columns
        result = obj.merge(result, how='outer', left_index=True, right_index=True).ix[:,-result.shape[1]:]
        result.columns = names
        return result

    def _define_paths(self, func, *args, **kwargs):
        if isinstance(func, compat.string_types):
            fast_path = lambda group: getattr(group, func)(*args, **kwargs)
            slow_path = lambda group: group.apply(
                lambda x: getattr(x, func)(*args, **kwargs), axis=self.axis)
        else:
            fast_path = lambda group: func(group, *args, **kwargs)
            slow_path = lambda group: group.apply(
                lambda x: func(x, *args, **kwargs), axis=self.axis)
        return fast_path, slow_path

    def _choose_path(self, fast_path, slow_path, group):
        path = slow_path
        res = slow_path(group)

        # if we make it here, test if we can use the fast path
        try:
            res_fast = fast_path(group)

            # compare that we get the same results
            if res.shape == res_fast.shape:
                res_r = res.values.ravel()
                res_fast_r = res_fast.values.ravel()
                mask = notnull(res_r)
            if (res_r[mask] == res_fast_r[mask]).all():
                path = fast_path

        except:
            pass
        return path, res

    def _transform_item_by_item(self, obj, wrapper):
        # iterate through columns
        output = {}
        inds = []
        for i, col in enumerate(obj):
            try:
                output[col] = self[col].transform(wrapper)
                inds.append(i)
            except Exception:
                pass

        if len(output) == 0: # pragma: no cover
            raise TypeError('Transform function invalid for data types')

        columns = obj.columns
        if len(output) < len(obj.columns):
            columns = columns.take(inds)

        return DataFrame(output, index=obj.index, columns=columns)

    def filter(self, func, dropna=True, *args, **kwargs):
        """
Return a copy of a DataFrame excluding elements from groups that
do not satisfy the boolean criterion specified by func.

Parameters
----------
f : function
Function to apply to each subframe. Should return True or False.
dropna : Drop groups that do not pass the filter. True by default;
if False, groups that evaluate False are filled with NaNs.

Notes
-----
Each subframe is endowed the attribute 'name' in case you need to know
which group you are working on.

Example
--------
>>> grouped = df.groupby(lambda x: mapping[x])
>>> grouped.filter(lambda x: x['A'].sum() + x['B'].sum() > 0)
"""
        from pandas.tools.merge import concat

        indices = []

        obj = self._selected_obj
        gen = self.grouper.get_iterator(obj, axis=self.axis)

        fast_path, slow_path = self._define_paths(func, *args, **kwargs)

        path = None
        for name, group in gen:
            object.__setattr__(group, 'name', name)

            if path is None:
                # Try slow path and fast path.
                try:
                    path, res = self._choose_path(fast_path, slow_path, group)
                except Exception: # pragma: no cover
                    res = fast_path(group)
                    path = fast_path
            else:
                res = path(group)

            def add_indices():
                indices.append(self._get_index(name))

            # interpret the result of the filter
            if isinstance(res, (bool, np.bool_)):
                if res:
                    add_indices()
            else:
                if getattr(res, 'ndim', None) == 1:
                    val = res.ravel()[0]
                    if val and notnull(val):
                        add_indices()
                else:

                    # in theory you could do .all() on the boolean result ?
                    raise TypeError("the filter must return a boolean result")

        filtered = self._apply_filter(indices, dropna)
        return filtered


class DataFrameGroupBy(NDFrameGroupBy):
    _apply_whitelist = _dataframe_apply_whitelist

    _block_agg_axis = 1

    def __getitem__(self, key):
        if self._selection is not None:
            raise Exception('Column(s) %s already selected' % self._selection)

        if isinstance(key, (list, tuple, Series, np.ndarray)):
            if len(self.obj.columns.intersection(key)) != len(key):
                bad_keys = list(set(key).difference(self.obj.columns))
                raise KeyError("Columns not found: %s"
                               % str(bad_keys)[1:-1])
            return DataFrameGroupBy(self.obj, self.grouper, selection=key,
                                    grouper=self.grouper,
                                    exclusions=self.exclusions,
                                    as_index=self.as_index)

        elif not self.as_index:
            if key not in self.obj.columns:
                raise KeyError("Column not found: %s" % key)
            return DataFrameGroupBy(self.obj, self.grouper, selection=key,
                                    grouper=self.grouper,
                                    exclusions=self.exclusions,
                                    as_index=self.as_index)

        else:
            if key not in self.obj:
                raise KeyError("Column not found: %s" % key)
            # kind of a kludge
            return SeriesGroupBy(self.obj[key], selection=key,
                                 grouper=self.grouper,
                                 exclusions=self.exclusions)

    def _wrap_generic_output(self, result, obj):
        result_index = self.grouper.levels[0]

        if result:
            if self.axis == 0:
                result = DataFrame(result, index=obj.columns,
                                   columns=result_index).T
            else:
                result = DataFrame(result, index=obj.index,
                                   columns=result_index)
        else:
            result = DataFrame(result)

        return result

    def _get_data_to_aggregate(self):
        obj = self._obj_with_exclusions
        if self.axis == 1:
            return obj.T._data, 1
        else:
            return obj._data, 1

    def _wrap_aggregated_output(self, output, names=None):
        agg_axis = 0 if self.axis == 1 else 1
        agg_labels = self._obj_with_exclusions._get_axis(agg_axis)

        output_keys = self._decide_output_index(output, agg_labels)

        if not self.as_index:
            result = DataFrame(output, columns=output_keys)
            group_levels = self.grouper.get_group_levels()
            zipped = zip(self.grouper.names, group_levels)

            for i, (name, labels) in enumerate(zipped):
                result.insert(i, name, labels)
            result = result.consolidate()
        else:
            index = self.grouper.result_index
            result = DataFrame(output, index=index, columns=output_keys)

        if self.axis == 1:
            result = result.T

        return result.convert_objects()

    def _wrap_agged_blocks(self, items, blocks):
        if not self.as_index:
            index = np.arange(blocks[0].values.shape[1])
            mgr = BlockManager(blocks, [items, index])
            result = DataFrame(mgr)

            group_levels = self.grouper.get_group_levels()
            zipped = zip(self.grouper.names, group_levels)

            for i, (name, labels) in enumerate(zipped):
                result.insert(i, name, labels)
            result = result.consolidate()
        else:
            index = self.grouper.result_index
            mgr = BlockManager(blocks, [items, index])
            result = DataFrame(mgr)

        if self.axis == 1:
            result = result.T

        return result.convert_objects()

    def _iterate_column_groupbys(self):
        for i, colname in enumerate(self._selected_obj.columns):
            yield colname, SeriesGroupBy(self._selected_obj.iloc[:, i],
                                         selection=colname,
                                         grouper=self.grouper,
                                         exclusions=self.exclusions)

    def _apply_to_column_groupbys(self, func):
        from pandas.tools.merge import concat
        return concat(
            (func(col_groupby) for _, col_groupby
             in self._iterate_column_groupbys()),
            keys=self._selected_obj.columns, axis=1)

from pandas.tools.plotting import boxplot_frame_groupby
DataFrameGroupBy.boxplot = boxplot_frame_groupby


class PanelGroupBy(NDFrameGroupBy):

    def _iterate_slices(self):
        if self.axis == 0:
            # kludge
            if self._selection is None:
                slice_axis = self._selected_obj.items
            else:
                slice_axis = self._selection_list
            slicer = lambda x: self._selected_obj[x]
        else:
            raise NotImplementedError

        for val in slice_axis:
            if val in self.exclusions:
                continue

            yield val, slicer(val)

    def aggregate(self, arg, *args, **kwargs):
        """
Aggregate using input function or dict of {column -> function}

Parameters
----------
arg : function or dict
Function to use for aggregating groups. If a function, must either
work when passed a Panel or when passed to Panel.apply. If
pass a dict, the keys must be DataFrame column names

Returns
-------
aggregated : Panel
"""
        if isinstance(arg, compat.string_types):
            return getattr(self, arg)(*args, **kwargs)

        return self._aggregate_generic(arg, *args, **kwargs)

    def _wrap_generic_output(self, result, obj):
        if self.axis == 0:
            new_axes = list(obj.axes)
            new_axes[0] = self.grouper.result_index
        elif self.axis == 1:
            x, y, z = obj.axes
            new_axes = [self.grouper.result_index, z, x]
        else:
            x, y, z = obj.axes
            new_axes = [self.grouper.result_index, y, x]

        result = Panel._from_axes(result, new_axes)

        if self.axis == 1:
            result = result.swapaxes(0, 1).swapaxes(0, 2)
        elif self.axis == 2:
            result = result.swapaxes(0, 2)

        return result

    def _aggregate_item_by_item(self, func, *args, **kwargs):
        obj = self._obj_with_exclusions
        result = {}

        if self.axis > 0:
            for item in obj:
                try:
                    itemg = DataFrameGroupBy(obj[item],
                                             axis=self.axis - 1,
                                             grouper=self.grouper)
                    result[item] = itemg.aggregate(func, *args, **kwargs)
                except (ValueError, TypeError):
                    raise
            new_axes = list(obj.axes)
            new_axes[self.axis] = self.grouper.result_index
            return Panel._from_axes(result, new_axes)
        else:
            raise NotImplementedError

    def _wrap_aggregated_output(self, output, names=None):
        raise NotImplementedError


class NDArrayGroupBy(GroupBy):
    pass


#----------------------------------------------------------------------
# Splitting / application


class DataSplitter(object):

    def __init__(self, data, labels, ngroups, axis=0):
        self.data = data
        self.labels = com._ensure_int64(labels)
        self.ngroups = ngroups

        self.axis = axis

    @cache_readonly
    def slabels(self):
        # Sorted labels
        return com.take_nd(self.labels, self.sort_idx, allow_fill=False)

    @cache_readonly
    def sort_idx(self):
        # Counting sort indexer
        return _algos.groupsort_indexer(self.labels, self.ngroups)[0]

    def __iter__(self):
        sdata = self._get_sorted_data()

        if self.ngroups == 0:
            raise StopIteration

        starts, ends = lib.generate_slices(self.slabels, self.ngroups)

        for i, (start, end) in enumerate(zip(starts, ends)):
            # Since I'm now compressing the group ids, it's now not "possible"
            # to produce empty slices because such groups would not be observed
            # in the data
            # if start >= end:
            # raise AssertionError('Start %s must be less than end %s'
            # % (str(start), str(end)))
            yield i, self._chop(sdata, slice(start, end))

    def _get_sorted_data(self):
        return self.data.take(self.sort_idx, axis=self.axis, convert=False)

    def _chop(self, sdata, slice_obj):
        return sdata.iloc[slice_obj]

    def apply(self, f):
        raise NotImplementedError


class ArraySplitter(DataSplitter):
    pass


class SeriesSplitter(DataSplitter):

    def _chop(self, sdata, slice_obj):
        return sdata._get_values(slice_obj).to_dense()


class FrameSplitter(DataSplitter):

    def __init__(self, data, labels, ngroups, axis=0):
        super(FrameSplitter, self).__init__(data, labels, ngroups, axis=axis)

    def fast_apply(self, f, names):
        # must return keys::list, values::list, mutated::bool
        try:
            starts, ends = lib.generate_slices(self.slabels, self.ngroups)
        except:
            # fails when all -1
            return [], True

        sdata = self._get_sorted_data()
        results, mutated = lib.apply_frame_axis0(sdata, f, names, starts, ends)

        return results, mutated

    def _chop(self, sdata, slice_obj):
        if self.axis == 0:
            return sdata.iloc[slice_obj]
        else:
            return sdata._slice(slice_obj, axis=1) # ix[:, slice_obj]


class NDFrameSplitter(DataSplitter):

    def __init__(self, data, labels, ngroups, axis=0):
        super(NDFrameSplitter, self).__init__(data, labels, ngroups, axis=axis)

        self.factory = data._constructor

    def _get_sorted_data(self):
        # this is the BlockManager
        data = self.data._data

        # this is sort of wasteful but...
        sorted_axis = data.axes[self.axis].take(self.sort_idx)
        sorted_data = data.reindex_axis(sorted_axis, axis=self.axis)

        return sorted_data

    def _chop(self, sdata, slice_obj):
        return self.factory(sdata.get_slice(slice_obj, axis=self.axis))


def get_splitter(data, *args, **kwargs):
    if isinstance(data, Series):
        klass = SeriesSplitter
    elif isinstance(data, DataFrame):
        klass = FrameSplitter
    else:
        klass = NDFrameSplitter

    return klass(data, *args, **kwargs)


#----------------------------------------------------------------------
# Misc utilities


def get_group_index(label_list, shape):
    """
For the particular label_list, gets the offsets into the hypothetical list
representing the totally ordered cartesian product of all possible label
combinations.
"""
    if len(label_list) == 1:
        return label_list[0]

    n = len(label_list[0])
    group_index = np.zeros(n, dtype=np.int64)
    mask = np.zeros(n, dtype=bool)
    for i in range(len(shape)):
        stride = np.prod([x for x in shape[i + 1:]], dtype=np.int64)
        group_index += com._ensure_int64(label_list[i]) * stride
        mask |= label_list[i] < 0

    np.putmask(group_index, mask, -1)
    return group_index

_INT64_MAX = np.iinfo(np.int64).max


def _int64_overflow_possible(shape):
    the_prod = long(1)
    for x in shape:
        the_prod *= long(x)

    return the_prod >= _INT64_MAX


def decons_group_index(comp_labels, shape):
    # reconstruct labels
    label_list = []
    factor = 1
    y = 0
    x = comp_labels
    for i in reversed(range(len(shape))):
        labels = (x - y) % (factor * shape[i]) // factor
        np.putmask(labels, comp_labels < 0, -1)
        label_list.append(labels)
        y = labels * factor
        factor *= shape[i]
    return label_list[::-1]


def _indexer_from_factorized(labels, shape, compress=True):
    if _int64_overflow_possible(shape):
        indexer = np.lexsort(np.array(labels[::-1]))
        return indexer

    group_index = get_group_index(labels, shape)

    if compress:
        comp_ids, obs_ids = _compress_group_index(group_index)
        max_group = len(obs_ids)
    else:
        comp_ids = group_index
        max_group = com._long_prod(shape)

    if max_group > 1e6:
        # Use mergesort to avoid memory errors in counting sort
        indexer = comp_ids.argsort(kind='mergesort')
    else:
        indexer, _ = _algos.groupsort_indexer(comp_ids.astype(np.int64),
                                              max_group)

    return indexer


def _lexsort_indexer(keys, orders=None, na_position='last'):
    labels = []
    shape = []
    if isinstance(orders, bool):
        orders = [orders] * len(keys)
    elif orders is None:
        orders = [True] * len(keys)

    for key, order in zip(keys, orders):
        key = np.asanyarray(key)
        rizer = _hash.Factorizer(len(key))

        if not key.dtype == np.object_:
            key = key.astype('O')

        # factorize maps nans to na_sentinel=-1
        ids = rizer.factorize(key, sort=True)
        n = len(rizer.uniques)
        mask = (ids == -1)
        if order: # ascending
            if na_position == 'last':
                ids = np.where(mask, n, ids)
            elif na_position == 'first':
                ids += 1
            else:
                raise ValueError('invalid na_position: {!r}'.format(na_position))
        else: # not order means descending
            if na_position == 'last':
                ids = np.where(mask, n, n-ids-1)
            elif na_position == 'first':
                ids = np.where(mask, 0, n-ids)
            else:
                raise ValueError('invalid na_position: {!r}'.format(na_position))
        if mask.any():
            n += 1
        shape.append(n)
        labels.append(ids)
    return _indexer_from_factorized(labels, shape)

def _nargsort(items, kind='quicksort', ascending=True, na_position='last'):
    """
This is intended to be a drop-in replacement for np.argsort which handles NaNs
It adds ascending and na_position parameters.
GH #6399, #5231
"""
    items = np.asanyarray(items)
    idx = np.arange(len(items))
    mask = isnull(items)
    non_nans = items[~mask]
    non_nan_idx = idx[~mask]
    nan_idx = np.nonzero(mask)[0]
    if not ascending:
        non_nans = non_nans[::-1]
        non_nan_idx = non_nan_idx[::-1]
    indexer = non_nan_idx[non_nans.argsort(kind=kind)]
    if not ascending:
        indexer = indexer[::-1]
    # Finally, place the NaNs at the end or the beginning according to na_position
    if na_position == 'last':
        indexer = np.concatenate([indexer, nan_idx])
    elif na_position == 'first':
        indexer = np.concatenate([nan_idx, indexer])
    else:
        raise ValueError('invalid na_position: {!r}'.format(na_position))
    return indexer


class _KeyMapper(object):

    """
Ease my suffering. Map compressed group id -> key tuple
"""

    def __init__(self, comp_ids, ngroups, labels, levels):
        self.levels = levels
        self.labels = labels
        self.comp_ids = comp_ids.astype(np.int64)

        self.k = len(labels)
        self.tables = [_hash.Int64HashTable(ngroups) for _ in range(self.k)]

        self._populate_tables()

    def _populate_tables(self):
        for labs, table in zip(self.labels, self.tables):
            table.map(self.comp_ids, labs.astype(np.int64))

    def get_key(self, comp_id):
        return tuple(level[table.get_item(comp_id)]
                     for table, level in zip(self.tables, self.levels))


def _get_indices_dict(label_list, keys):
    shape = [len(x) for x in keys]
    group_index = get_group_index(label_list, shape)

    sorter, _ = _algos.groupsort_indexer(com._ensure_int64(group_index),
                                         np.prod(shape))

    sorter_int = com._ensure_platform_int(sorter)

    sorted_labels = [lab.take(sorter_int) for lab in label_list]
    group_index = group_index.take(sorter_int)

    return lib.indices_fast(sorter, group_index, keys, sorted_labels)


#----------------------------------------------------------------------
# sorting levels...cleverly?


def _compress_group_index(group_index, sort=True):
    """
Group_index is offsets into cartesian product of all possible labels. This
space can be huge, so this function compresses it, by computing offsets
(comp_ids) into the list of unique labels (obs_group_ids).
"""

    table = _hash.Int64HashTable(min(1000000, len(group_index)))

    group_index = com._ensure_int64(group_index)

    # note, group labels come out ascending (ie, 1,2,3 etc)
    comp_ids, obs_group_ids = table.get_labels_groupby(group_index)

    if sort and len(obs_group_ids) > 0:
        obs_group_ids, comp_ids = _reorder_by_uniques(obs_group_ids, comp_ids)

    return comp_ids, obs_group_ids


def _reorder_by_uniques(uniques, labels):
    # sorter is index where elements ought to go
    sorter = uniques.argsort()

    # reverse_indexer is where elements came from
    reverse_indexer = np.empty(len(sorter), dtype=np.int64)
    reverse_indexer.put(sorter, np.arange(len(sorter)))

    mask = labels < 0

    # move labels to right locations (ie, unsort ascending labels)
    labels = com.take_nd(reverse_indexer, labels, allow_fill=False)
    np.putmask(labels, mask, -1)

    # sort observed ids
    uniques = com.take_nd(uniques, sorter, allow_fill=False)

    return uniques, labels


_func_table = {
    builtins.sum: np.sum
}


_cython_table = {
    builtins.sum: 'sum',
    np.sum: 'sum',
    np.mean: 'mean',
    np.prod: 'prod',
    np.std: 'std',
    np.var: 'var',
    np.median: 'median',
    np.max: 'max',
    np.min: 'min'
}


def _intercept_function(func):
    return _func_table.get(func, func)


def _intercept_cython(func):
    return _cython_table.get(func)


def _groupby_indices(values):
    return _algos.groupby_indices(com._ensure_object(values))


def numpy_groupby(data, labels, axis=0):
    s = np.argsort(labels)
    keys, inv = np.unique(labels, return_inverse=True)
    i = inv.take(s)
    groups_at = np.where(i != np.concatenate(([-1], i[:-1])))[0]
    ordered_data = data.take(s, axis=axis)
    group_sums = np.add.reduceat(ordered_data, groups_at, axis=axis)

    return group_sums
Something went wrong with that request. Please try again.