More parser gliches

Summary:
Interesting programs and their outputs:

<?php
//?????>hello
---
hello

<?php
$foo->if();
---
<works>

<?php
$foo-> if()
---
<syntax error>

Anyway, PHP's parser is a huge mess.

Reviewed By: no one

Test Plan: Disabled "maybe_xhp" check, then `find . | grep php$ | xargs -n1 php -l`. No errors!

Revert Plan: ok

ext.cpp

@@ -100,9 +100,7 @@ static zend_op_array* xhp_compile_file(zend_file_handle* f, int type TSRMLS_DC)
         maybe_xhp = 1;
         break;
       }
-    } else if (*ii == 'e' && ii[1] == 'l' && ii[2] == 'e' && ii[3] == 'm' && ii[4] == 'e' && ii[5] == 'n' && ii[6] == 't') {
-//    } else if (memcmp(ii, "element", 7)) {
-      // why is this faster than memcmp? i'm bad at computers and i don't know.
+    } else if (!memcmp(ii, "element", 7)) {
       maybe_xhp = 1;
       break;
     }

xhp_parser.y

@@ -39,6 +39,7 @@
 %parse-param { code_rope* root }
 %lex-param { void* yyscanner }
 %error-verbose
+%expect 2 // i hate php's if syntax
 
 // Keywords
 %token BOGUS
@@ -154,7 +155,7 @@ statement:
     $$ = "{" + $2 + cr("}");
   }
 | function_declaration
-| if_statement // TODO
+| if_statement
 | while_statement
 | for_statement
 | foreach_statement
@@ -187,11 +188,32 @@ if_statement:
   t_IF t_LPAREN expression t_RPAREN statement {
     $$ = "if (" + $3 + ") " + $5;
   }
-| statement t_ELSE statement {
+| if_statement t_ELSE statement {
     $$ = $1 + " else " + $3;
   }
-| statement t_ELSEIF t_LPAREN expression t_RPAREN statement {
-    $$ = $1 + "elseif (" + $4 + ") " + $6;
+| if_statement t_ELSEIF t_LPAREN expression t_RPAREN statement {
+    $$ = $1 + " elseif (" + $4 + ") " + $6;
+  }
+| t_IF t_LPAREN expression t_RPAREN t_COLON statement_list elseif_list else_single t_ENDIF semicolon {
+    $$ = "if (" + $3 + "): " + $6 + $7 + $8 + " endif" + $10;
+  }
+;
+
+elseif_list:
+  /* empty */ {
+    $$ = "";
+  }
+| elseif_list t_ELSEIF t_LPAREN expression t_RPAREN t_COLON statement_list {
+    $$ = $1 + " elseif(" + $4 + "):" + $7;
+  }
+;
+
+else_single:
+  /* empty */ {
+    $$ = "";
+  }
+| t_ELSE statement_list {
+    $$ = " else: " + $2;
   }
 ;
 
@@ -430,6 +452,10 @@ expression:
 | t_ARRAY identifier {
     $$ = "array " + $2;
   }
+| t_ARRAY t_BIT_AND identifier {
+    // handles function foo(array &bar){}
+    $$ = "array &" + $3;
+  }
 | t_ARRAY t_LPAREN array_pair_list t_RPAREN {
     $$ = "array(" + $3 + ")";
   }
@@ -566,7 +592,8 @@ expression:
     $$ = $1 + "::" + $3;
   }
 | expression t_ARROW expression {
-    $$ = $1 + "->" + $3;
+    // Need -> in the rope because "$foo-> if()" is a syntax error, but if you remove the space it's not.
+    $$ = $1 + cr("->") + $3;
   }
 | t_INCR expression {
     $$ = "++" + $2;

xhp_scanner.l

@@ -83,10 +83,39 @@ B b?
 }
 <PHP,PHP_NO_RESERVED_WORDS,XHP_ATTR>{
   [ \t\x0b\x0c\xa0]+ /* eat it up */
-  ("//"|"#").*{NEWLINE} |
   {NEWLINE} {
     ++yylloc->internal_line;
   }
+  "//"|"#" {
+    char c;
+    for (;;) {
+      while ((c = yyinput(yyscanner)) != '?' && c != '\r' && c != '\n' && c != EOF);
+      honk:
+      if (c == '\r') {
+        ++yylloc->internal_line;
+        c == yyinput(yyscanner);
+        if (c != '\n') {
+          unput(c);
+          break;
+        }
+      } else if (c == '\n') {
+        ++yylloc->internal_line;
+        break;
+      }
+      if (c == '?') {
+        c = yyinput(yyscanner);
+        if (c == '>') {
+          unput('>');
+          unput('?');
+          break;
+        }
+        goto honk;
+      }
+      if (c == EOF) {
+        return tok(0);
+      }
+    }
+  }
   "/*" {
     char c;
     for (;;) {
@@ -133,7 +162,13 @@ B b?
   (?i:default) return tok(t_DEFAULT);
   (?i:break) return tok(t_BREAK);
   (?i:continue) return tok(t_CONTINUE);
-  (?i:function) return tok(t_FUNCTION);
+  (?i:function) {
+    int tt = tok(t_FUNCTION);
+    // needed so you can say function element(){}
+    // this only works on objects, though.
+    flex_push_state(PHP_NO_RESERVED_WORDS, yyg);
+    return tt;
+  }
   (?i:const) return tok(t_CONST);
   (?i:instanceof) return tok(t_INSTANCEOF);
   (?i:return) return tok(t_RETURN);
@@ -195,7 +230,11 @@ B b?
     return tt;
   }
   "=>" return tok(t_DOUBLE_ARROW);
-  "::" return tok(t_HEBREW_THING);
+  "::" {
+    int tt = tok(t_HEBREW_THING);
+    flex_push_state(PHP_NO_RESERVED_WORDS, yyg);
+    return tt;
+  }
   "." return tok(t_CONCAT);
   "," return tok(t_COMMA);
   ";" return tok(t_SEMICOLON);
@@ -262,11 +301,11 @@ B b?
     *yylval = yytext;
     return tok(t_NUMBER);
   }
-  [0-9]*\.?[0-9]+[eE]-?[0-9]{1,3}  {
+  [0-9]*\.?[0-9]+[eE][\-+]?[0-9]{1,3}  {
     *yylval = yytext;
     return tok(t_NUMBER);
   }
-  [0-9]+ |
+  [0-9]+\.? |
   [0-9]*\.[0-9]+ {
     *yylval = yytext;
     return tok(t_NUMBER);
@@ -281,7 +320,7 @@ B b?
     eom += 3;
     while (*eom == '\t' || *eom == ' ') ++eom;
     size_t len = yyleng - (eom - yytext) - 1;
-    if (eom[len - 2] == '\r') { // windows [\r\n] encoding
+    if (eom[len - 1] == '\r') { // windows [\r\n] encoding
       --len;
     }
     yyextra->heredoc_eom = eom;
@@ -509,6 +548,6 @@ static int flex_tok(int t, void* y) {
   if (YY_START == PHP_NO_RESERVED_WORDS) {
     flex_pop_state(yyg);
   }
-//  printf("\ntok:%s\n", yytokname(t));
+//  fprintf(stderr, "\ntok:%s\n", yytokname(t));
   return t;
 }