Skip to content
This repository

add sugar methods/properties to AsyncResult #1548

Merged
merged 8 commits into from about 2 years ago

2 participants

Min RK Fernando Perez
Min RK
Owner
minrk commented

Adds the following attributes / methods:

  • ar.wall_time = received - submitted
  • ar.serial_time = sum of serial computation time
  • ar.elapsed = time since submission (wall_time if done)
  • ar.progress = (int) number of sub-tasks that have completed
  • len(ar) = # of tasks
  • ar.wait_interactive(): prints progress

These are simple methods derived from the metadata/timestamps already created. But I've been persuaded by @wesm's practice of including simple methods that do useful (and/or cool) things.

This also required/revealed some minor fixes/cleanup to clear_output in some cases:

  • dedent base core.displaypub.clear_output, so it's actually defined in the class
  • clear_output publishes '\r\b', so it will clear terminal-like frontends that don't understand full clear_output behavior.
  • core.display.clear_output() still works, even outside an IPython session.
Min RK
Owner
minrk commented

At @fperez request, I had a go at addressing the flicker when using clear_output in the notebook. I used a simple timeout, which gets cleared/flushed immediately on the next output-related action.

I won't object to moving that to a separate PR, but I did it here as it is related, and this new code makes significant use of the clear;print;repeat pattern for which the flicker is most problematic.

Min RK
Owner
minrk commented

this notebook shows a few examples of print, the new AR.wait_interactive, and plotting with clear_output, all of which seem better behaved than previously.

Min RK minrk referenced this pull request
Merged

clear_output improvements #1563

Min RK
Owner
minrk commented

This PR now depends on PR #1563

Fernando Perez
Owner

@minrk, I merged #1563 but now it's producing some conflicts. Could you rebase this one? I'll be happy to review it quickly...

Fernando Perez
Owner

And as I commented in that other one, might as well make the notebook with examples part of the PR, so we beef up the notebooks we provide with the system for users. That will become even more useful once we start including their html for in the docs, along with a link to the original .ipynb file, as we'll be able to point users to those pages in the docs.

Min RK add sugar methods/properties to AsyncResult
* ar.wall_time = received - submitted
* ar.serial_time = sum of serial computation time
* ar.elapsed = time since submission (wall_time if done)
* ar.progress = (int) number of sub-tasks that have completed
* ar.wait_interactive(): prints progress
* len(ar) = # of tasks
148c775
Min RK
Owner
minrk commented

rebased - I'll add the notebook in the morning, probably.

Min RK
Owner
minrk commented

Progress notebook added, demos:

  • clear_output/print/flush
  • AsyncResult.wait_interactive
  • clear_output/display(Figure)
  • HTML/JS progress bar
  • PyMC ProgressBar class
docs/examples/notebooks/Progress.ipynb
((19 lines not shown))
  19
+      "What the notebook *does* support is explicit `clear_output`, and you can use this to replace previous",
  20
+      "output (specifying stdout/stderr or the special IPython display outputs)."
  21
+     ]
  22
+    },
  23
+    {
  24
+     "cell_type": "markdown",
  25
+     "source": [
  26
+      "A simple example printing our progress iterating through a list:"
  27
+     ]
  28
+    },
  29
+    {
  30
+     "cell_type": "code",
  31
+     "collapsed": false,
  32
+     "input": [
  33
+      "from IPython.core.display import clear_output",
  34
+      "for i in range(10):",
2
Fernando Perez Owner
fperez added a note

It's missing an import time here.

Min RK Owner
minrk added a note

ah, yes. I have import os,sys,time in my startup files, so I always forget to import them.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
docs/examples/notebooks/Progress.ipynb
((76 lines not shown))
  76
+    },
  77
+    {
  78
+     "cell_type": "code",
  79
+     "collapsed": false,
  80
+     "input": [
  81
+      "%pylab inline"
  82
+     ],
  83
+     "language": "python",
  84
+     "outputs": [],
  85
+     "prompt_number": 6
  86
+    },
  87
+    {
  88
+     "cell_type": "code",
  89
+     "collapsed": false,
  90
+     "input": [
  91
+      "from scipy.special import jn",
3
Fernando Perez Owner
fperez added a note

I'd rewrite this cell as:

from scipy.special import jn
x = linspace(0,5)
f, ax = plt.subplots()
ax.set_title("Bessel functions")

for n in range(1,10):
    time.sleep(1)
    ax.plot(x, jn(x,n))
    clear_output()
    display(f)

# close the figure at the end, so we don't get a duplicate
# of the last plot
plt.close()

It's a bit cleaner api-wise regarding matplotlib, and it completely eliminates the flicker. The trick is to do all plotting with the axis object, which doesn't force a draw call. Then, we can do clear_output right before the display call, and most likely (unless it's a super-complex plot), it will finish before any flickering happens.

You can mention this and update the comment below accordingly, indicating how this will eliminate completely all flickering in most common cases (though it can still appear for sufficiently complex plots).

Min RK Owner
minrk added a note

Thanks, I made this change, though it isn't true that it eliminates the flicker. Actually eliminating flicker will require changing how clear_output works in a more complicated way (don't resize height until after new content is drawn), because as it is now, it's still resizing to zero immediately before drawing the image, which takes more than zero time (though it is very quick).

Min RK Owner
minrk added a note

Hm, nevermind - the flicker does seem to be gone. That means that I misunderstood what was causing the flicker, and also suggests that calls to plot() results in some IOPub message, possibly an empty sys.stdout.flush()?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
docs/examples/notebooks/Progress.ipynb
... ...
@@ -0,0 +1,226 @@
  1
+{
  2
+ "metadata": {
  3
+  "name": "Progress"
  4
+ },
  5
+ "nbformat": 3,
  6
+ "worksheets": [
  7
+  {
  8
+   "cells": [
  9
+    {
  10
+     "cell_type": "markdown",
  11
+     "source": [
  12
+      "# Progress bars and clear_output",
1
Fernando Perez Owner
fperez added a note

Perhaps change the title to 'Simple animations, progress bars and output area cleaning'? I'd also suggest renaming the file to something slightly more descriptive, say animations_progress_bars or similar (casing up to you). 'Progress' by itself as filename is a bit cryptic.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Fernando Perez fperez commented on the diff
IPython/parallel/client/asyncresult.py
... ...
@@ -255,7 +259,76 @@ def __iter__(self):
255 259
             # already done
256 260
             for r in rlist:
257 261
                 yield r
258  
-
  262
+    
  263
+    def __len__(self):
  264
+        return len(self.msg_ids)
  265
+    
  266
+    #-------------------------------------
  267
+    # Sugar methods and attributes
2
Fernando Perez Owner
fperez added a note

This is fantastic, and the implementation looks great. I just think it needs mentioning somewhere in the docs, though... Otherwise users are never going to find out about these things. It could be just a paragraph pointing out their existence, but I think we do need something in the docs about these guys.

Min RK Owner
minrk added a note

I've been a bit reluctant to add things to the current docs, because I really want to trash them entirely and start from scratch, but obviously I should document new features as they are added.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Fernando Perez
Owner

Minor comments inline, easy to fix up and we'll be good to go. This is awesome.

Min RK add Animations and Progress example notebook
demos intermediate progress with:

* clear_output/print/flush
* AsyncResult.wait_interactive
* clear_output/display(Figure)
* HTML/JS progress bar
* PyMC ProgressBar class
f5e3cc0
Min RK
Owner
minrk commented

I made your edits to the notebook, and started a doc for the AsyncResult object. I really want to make it clear that the AsyncResult object itself is where most of our API lives. At least the nifty parts.

I have one question we should resolve before merging. In wall_time, I use received - submitted, which is the actual roundtrip time for the Client. The only problem with this one is that the received timestamp is not especially reliable, particularly in interactive use cases. The reason being that this timestamp is made when the result is pulled off of the queue in the Client, as a result of Client.spin(). So if the result of the computation arrives either while the user is sitting idle at an interactive prompt or just performing some client-side computation, that time is added to the wall_time measure.

This is a larger issue for AsyncHubResults (those fetched from the Hub), where the received stamp could technically be days after the computation finished.

Possible partial solutions to this problem:

  • HubResults should use completed-submitted, ignoring result-reply overhead
  • all AsyncResults should use completed-submitted (most consistent, but probably not most useful/interesting)
  • HubResults should just not have a wall_time property
  • Add an analogue to received to the Hub's DB, and use that in HubResults
  • Leave it as-is, and let users deal with wall_time rarely being useful in HubResults (it would still be accurate if the HubResult is requested and waited upon while the computation is pending)

I'm personally inclined towards either one of the last two, but I want there to be a convenient

There are a couple of other timings that might be useful to add:

  • last completed - first started (actual time spent working, excluding overhead - I don't know what it's name should be)
  • last completed - first submitted (wall_time excluding reply overhead, which may not be interesting or meaningful)
  • last received - first started (excludes start overhead, which could have just been waiting for other jobs in the queue)
  • idle time (total time spent on each engine not during a computation) - this might be computed per-engine to draw one of those pipeline concurrency figures.

I don't know how many of these I should actually include. Perhaps I should just make a method that makes it easy to do any of the last/first X - first/last Y deltas.

Fernando Perez

Ouch, no... Let's avoid filenames with spaces in them in the repo, they always end up causing trouble for automated handling. I know the underscores look ugly in files that are much more 'gui' oriented, but spaces in filenames really are a timebomb waiting to happen for scripted work.

Owner

Seriously? The notebook is a GUI tool, and GUI files should have reasonable names. Underscores are a gross hack to appease pathetic tools.

Owner

But managing directories with filenames that have spaces in them with automated tools is extraordinarily error prone. I don't want to end up having to debug one day a script for doc generation (for example) because something downstream choked on a space. I've been bitten by that particular problem enough times in my life that I simply don't want to waste my time with that kind of problem. Practicality beats purity, in this case...

Owner

I don't agree at all, and have much less patience or sympathy for shitty tools, but I will make the change.

I would think that handling simple sensible filenames is a reasonable expectation in this century.

Owner

If you really feel strongly about it, leave it as-is. If we ever run into problems I'll just rename things outright; I'm not trying to upset you over a minor change, just trying to avoid possible annoyances down the road.

Owner

It's not a big deal - I just get annoyed at how bad command-line tools are sometimes, and how *ix standard practices like this one tend to be based on how these tools suck at basic functionality.

I do think that if this is to be standard practice in the notebook, then we should proabably unescape underscores, and draw regular spaces in the notebook list. A listing full of underscores that are really spaces is pretty ugly, and simply doesn't fit in with the web today, where spaces are handled frequently and without issue.

added some commits
Min RK add 'received' timestamp to DB
allows 'wall_time' to make sense in cases other than simple waiting AsyncResult.
bd8a8ec
Min RK add AsyncResult.timedelta
for computing various comparisons of timestamps in AsyncResults
d004201
Min RK remove spaces in progress notebook filename
appease crappy tools that can't deal with reasonable filenames.
978ac5b
Min RK test new AsyncResult properties 66cc8f8
Min RK document new AsyncResult properties 89a00db
Min RK
Owner
minrk commented

Okay, review addressed I think:

  • spaces removed from filename
  • new properties documented
  • some basic tests added
  • I went with adding received to the Hub's DB for resolving the wall_time issue for HubResults
  • I added AsyncResult.timedelta() for comparing a pair of timestamp sets, which is used in AR.wall_time.

Did I miss anything else?

Fernando Perez

awesome, thanks for all these tests!

Fernando Perez

And this is also great.

Min RK add Client.spin_thread()
runs Client.spin() in a background thread at a set interval
472cd24
Min RK
Owner
minrk commented

Added Client.spin_thread(interval) / stop_spin_thread() for running spin in a background thread, to keep zmq queue clear.

Fernando Perez
Owner

Great, merging now. Awesome job, thanks!

Fernando Perez fperez merged commit e0b4311 into from
Fernando Perez fperez closed this
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Showing 8 unique commits by 1 author.

Apr 13, 2012
Min RK add sugar methods/properties to AsyncResult
* ar.wall_time = received - submitted
* ar.serial_time = sum of serial computation time
* ar.elapsed = time since submission (wall_time if done)
* ar.progress = (int) number of sub-tasks that have completed
* ar.wait_interactive(): prints progress
* len(ar) = # of tasks
148c775
Apr 14, 2012
Min RK add Animations and Progress example notebook
demos intermediate progress with:

* clear_output/print/flush
* AsyncResult.wait_interactive
* clear_output/display(Figure)
* HTML/JS progress bar
* PyMC ProgressBar class
f5e3cc0
Min RK add 'received' timestamp to DB
allows 'wall_time' to make sense in cases other than simple waiting AsyncResult.
bd8a8ec
Min RK add AsyncResult.timedelta
for computing various comparisons of timestamps in AsyncResults
d004201
Min RK remove spaces in progress notebook filename
appease crappy tools that can't deal with reasonable filenames.
978ac5b
Min RK test new AsyncResult properties 66cc8f8
Min RK document new AsyncResult properties 89a00db
Min RK add Client.spin_thread()
runs Client.spin() in a background thread at a set interval
472cd24
This page is out of date. Refresh to see the latest.
111  IPython/parallel/client/asyncresult.py
@@ -15,13 +15,17 @@
15 15
 # Imports
16 16
 #-----------------------------------------------------------------------------
17 17
 
  18
+import sys
18 19
 import time
  20
+from datetime import datetime
19 21
 
20 22
 from zmq import MessageTracker
21 23
 
  24
+from IPython.core.display import clear_output
22 25
 from IPython.external.decorator import decorator
23 26
 from IPython.parallel import error
24 27
 
  28
+
25 29
 #-----------------------------------------------------------------------------
26 30
 # Classes
27 31
 #-----------------------------------------------------------------------------
@@ -255,7 +259,112 @@ def __iter__(self):
255 259
             # already done
256 260
             for r in rlist:
257 261
                 yield r
258  
-
  262
+    
  263
+    def __len__(self):
  264
+        return len(self.msg_ids)
  265
+    
  266
+    #-------------------------------------
  267
+    # Sugar methods and attributes
  268
+    #-------------------------------------
  269
+    
  270
+    def timedelta(self, start, end, start_key=min, end_key=max):
  271
+        """compute the difference between two sets of timestamps
  272
+        
  273
+        The default behavior is to use the earliest of the first
  274
+        and the latest of the second list, but this can be changed
  275
+        by passing a different
  276
+        
  277
+        Parameters
  278
+        ----------
  279
+        
  280
+        start : one or more datetime objects (e.g. ar.submitted)
  281
+        end : one or more datetime objects (e.g. ar.received)
  282
+        start_key : callable
  283
+            Function to call on `start` to extract the relevant
  284
+            entry [defalt: min]
  285
+        end_key : callable
  286
+            Function to call on `end` to extract the relevant
  287
+            entry [default: max]
  288
+        
  289
+        Returns
  290
+        -------
  291
+        
  292
+        dt : float
  293
+            The time elapsed (in seconds) between the two selected timestamps.
  294
+        """
  295
+        if not isinstance(start, datetime):
  296
+            # handle single_result AsyncResults, where ar.stamp is single object,
  297
+            # not a list
  298
+            start = start_key(start)
  299
+        if not isinstance(end, datetime):
  300
+            # handle single_result AsyncResults, where ar.stamp is single object,
  301
+            # not a list
  302
+            end = end_key(end)
  303
+        return (end - start).total_seconds()
  304
+        
  305
+    @property
  306
+    def progress(self):
  307
+        """the number of tasks which have been completed at this point.
  308
+        
  309
+        Fractional progress would be given by 1.0 * ar.progress / len(ar)
  310
+        """
  311
+        self.wait(0)
  312
+        return len(self) - len(set(self.msg_ids).intersection(self._client.outstanding))
  313
+    
  314
+    @property
  315
+    def elapsed(self):
  316
+        """elapsed time since initial submission"""
  317
+        if self.ready():
  318
+            return self.wall_time
  319
+        
  320
+        now = submitted = datetime.now()
  321
+        for msg_id in self.msg_ids:
  322
+            if msg_id in self._client.metadata:
  323
+                stamp = self._client.metadata[msg_id]['submitted']
  324
+                if stamp and stamp < submitted:
  325
+                    submitted = stamp
  326
+        return (now-submitted).total_seconds()
  327
+    
  328
+    @property
  329
+    @check_ready
  330
+    def serial_time(self):
  331
+        """serial computation time of a parallel calculation
  332
+        
  333
+        Computed as the sum of (completed-started) of each task
  334
+        """
  335
+        t = 0
  336
+        for md in self._metadata:
  337
+            t += (md['completed'] - md['started']).total_seconds()
  338
+        return t
  339
+    
  340
+    @property
  341
+    @check_ready
  342
+    def wall_time(self):
  343
+        """actual computation time of a parallel calculation
  344
+        
  345
+        Computed as the time between the latest `received` stamp
  346
+        and the earliest `submitted`.
  347
+        
  348
+        Only reliable if Client was spinning/waiting when the task finished, because
  349
+        the `received` timestamp is created when a result is pulled off of the zmq queue,
  350
+        which happens as a result of `client.spin()`.
  351
+        
  352
+        For similar comparison of other timestamp pairs, check out AsyncResult.timedelta.
  353
+        
  354
+        """
  355
+        return self.timedelta(self.submitted, self.received)
  356
+    
  357
+    def wait_interactive(self, interval=1., timeout=None):
  358
+        """interactive wait, printing progress at regular intervals"""
  359
+        N = len(self)
  360
+        tic = time.time()
  361
+        while not self.ready() and (timeout is None or time.time() - tic <= timeout):
  362
+            self.wait(interval)
  363
+            clear_output()
  364
+            print "%4i/%i tasks finished after %4i s" % (self.progress, N, self.elapsed),
  365
+            sys.stdout.flush()
  366
+        print
  367
+        print "done"
259 368
 
260 369
 
261 370
 class AsyncMapResult(AsyncResult):
55  IPython/parallel/client/client.py
@@ -18,6 +18,7 @@
18 18
 import os
19 19
 import json
20 20
 import sys
  21
+from threading import Thread, Event
21 22
 import time
22 23
 import warnings
23 24
 from datetime import datetime
@@ -36,7 +37,7 @@
36 37
 from IPython.utils.localinterfaces import LOCAL_IPS
37 38
 from IPython.utils.path import get_ipython_dir
38 39
 from IPython.utils.traitlets import (HasTraits, Integer, Instance, Unicode,
39  
-                                    Dict, List, Bool, Set)
  40
+                                    Dict, List, Bool, Set, Any)
40 41
 from IPython.external.decorator import decorator
41 42
 from IPython.external.ssh import tunnel
42 43
 
@@ -249,6 +250,8 @@ class Client(HasTraits):
249 250
     metadata = Instance('collections.defaultdict', (Metadata,))
250 251
     history = List()
251 252
     debug = Bool(False)
  253
+    _spin_thread = Any()
  254
+    _stop_spinning = Any()
252 255
 
253 256
     profile=Unicode()
254 257
     def _profile_default(self):
@@ -299,6 +302,7 @@ def __init__(self, url_or_file=None, profile=None, profile_dir=None, ipython_dir
299 302
         if context is None:
300 303
             context = zmq.Context.instance()
301 304
         self._context = context
  305
+        self._stop_spinning = Event()
302 306
 
303 307
         self._setup_profile_dir(self.profile, profile_dir, ipython_dir)
304 308
         if self._cd is not None:
@@ -785,12 +789,59 @@ def ids(self):
785 789
     def close(self):
786 790
         if self._closed:
787 791
             return
  792
+        self.stop_spin_thread()
788 793
         snames = filter(lambda n: n.endswith('socket'), dir(self))
789 794
         for socket in map(lambda name: getattr(self, name), snames):
790 795
             if isinstance(socket, zmq.Socket) and not socket.closed:
791 796
                 socket.close()
792 797
         self._closed = True
793 798
 
  799
+    def _spin_every(self, interval=1):
  800
+        """target func for use in spin_thread"""
  801
+        while True:
  802
+            if self._stop_spinning.is_set():
  803
+                return
  804
+            time.sleep(interval)
  805
+            self.spin()
  806
+
  807
+    def spin_thread(self, interval=1):
  808
+        """call Client.spin() in a background thread on some regular interval
  809
+        
  810
+        This helps ensure that messages don't pile up too much in the zmq queue
  811
+        while you are working on other things, or just leaving an idle terminal.
  812
+        
  813
+        It also helps limit potential padding of the `received` timestamp
  814
+        on AsyncResult objects, used for timings.
  815
+        
  816
+        Parameters
  817
+        ----------
  818
+        
  819
+        interval : float, optional
  820
+            The interval on which to spin the client in the background thread
  821
+            (simply passed to time.sleep).
  822
+        
  823
+        Notes
  824
+        -----
  825
+        
  826
+        For precision timing, you may want to use this method to put a bound
  827
+        on the jitter (in seconds) in `received` timestamps used
  828
+        in AsyncResult.wall_time.
  829
+        
  830
+        """
  831
+        if self._spin_thread is not None:
  832
+            self.stop_spin_thread()
  833
+        self._stop_spinning.clear()
  834
+        self._spin_thread = Thread(target=self._spin_every, args=(interval,))
  835
+        self._spin_thread.daemon = True
  836
+        self._spin_thread.start()
  837
+    
  838
+    def stop_spin_thread(self):
  839
+        """stop background spin_thread, if any"""
  840
+        if self._spin_thread is not None:
  841
+            self._stop_spinning.set()
  842
+            self._spin_thread.join()
  843
+            self._spin_thread = None
  844
+
794 845
     def spin(self):
795 846
         """Flush any registration notifications and execution results
796 847
         waiting in the ZMQ queue.
@@ -1277,6 +1328,8 @@ def result_status(self, msg_ids, status_only=True):
1277 1328
 
1278 1329
                 md = self.metadata[msg_id]
1279 1330
                 md.update(self._extract_metadata(header, parent, rcontent))
  1331
+                if rec.get('received'):
  1332
+                    md['received'] = rec['received']
1280 1333
                 md.update(iodict)
1281 1334
 
1282 1335
                 if rcontent['status'] == 'ok':
9  IPython/parallel/controller/hub.py
@@ -63,6 +63,7 @@ def empty_record():
63 63
         'started': None,
64 64
         'completed': None,
65 65
         'resubmitted': None,
  66
+        'received': None,
66 67
         'result_header' : None,
67 68
         'result_content' : None,
68 69
         'result_buffers' : None,
@@ -88,6 +89,7 @@ def init_record(msg):
88 89
         'started': None,
89 90
         'completed': None,
90 91
         'resubmitted': None,
  92
+        'received': None,
91 93
         'result_header' : None,
92 94
         'result_content' : None,
93 95
         'result_buffers' : None,
@@ -630,6 +632,7 @@ def save_queue_result(self, idents, msg):
630 632
         result = {
631 633
             'result_header' : rheader,
632 634
             'result_content': msg['content'],
  635
+            'received': datetime.now(),
633 636
             'started' : started,
634 637
             'completed' : completed
635 638
         }
@@ -732,7 +735,8 @@ def save_task_result(self, idents, msg):
732 735
                 'result_content': msg['content'],
733 736
                 'started' : started,
734 737
                 'completed' : completed,
735  
-                'engine_uuid': engine_uuid
  738
+                'received' : datetime.now(),
  739
+                'engine_uuid': engine_uuid,
736 740
             }
737 741
 
738 742
             result['result_buffers'] = msg['buffers']
@@ -1176,11 +1180,12 @@ def finish(reply):
1176 1180
     def _extract_record(self, rec):
1177 1181
         """decompose a TaskRecord dict into subsection of reply for get_result"""
1178 1182
         io_dict = {}
1179  
-        for key in 'pyin pyout pyerr stdout stderr'.split():
  1183
+        for key in ('pyin', 'pyout', 'pyerr', 'stdout', 'stderr'):
1180 1184
                 io_dict[key] = rec[key]
1181 1185
         content = { 'result_content': rec['result_content'],
1182 1186
                             'header': rec['header'],
1183 1187
                             'result_header' : rec['result_header'],
  1188
+                            'received' : rec['received'],
1184 1189
                             'io' : io_dict,
1185 1190
                           }
1186 1191
         if rec['result_buffers']:
3  IPython/parallel/controller/sqlitedb.py
@@ -117,6 +117,7 @@ class SQLiteDB(BaseDB):
117 117
             'started',
118 118
             'completed',
119 119
             'resubmitted',
  120
+            'received',
120 121
             'result_header' ,
121 122
             'result_content' ,
122 123
             'result_buffers' ,
@@ -138,6 +139,7 @@ class SQLiteDB(BaseDB):
138 139
             'started' : 'timestamp',
139 140
             'completed' : 'timestamp',
140 141
             'resubmitted' : 'timestamp',
  142
+            'received' : 'timestamp',
141 143
             'result_header' : 'dict text',
142 144
             'result_content' : 'dict text',
143 145
             'result_buffers' : 'bufs blob',
@@ -245,6 +247,7 @@ def _init_db(self):
245 247
                 started timestamp,
246 248
                 completed timestamp,
247 249
                 resubmitted timestamp,
  250
+                received timestamp,
248 251
                 result_header dict text,
249 252
                 result_content dict text,
250 253
                 result_buffers bufs blob,
82  IPython/parallel/tests/test_asyncresult.py
@@ -16,10 +16,11 @@
16 16
 # Imports
17 17
 #-------------------------------------------------------------------------------
18 18
 
  19
+import time
19 20
 
20 21
 from IPython.parallel.error import TimeoutError
21 22
 
22  
-from IPython.parallel import error
  23
+from IPython.parallel import error, Client
23 24
 from IPython.parallel.tests import add_engines
24 25
 from .clienttest import ClusterTestCase
25 26
 
@@ -121,5 +122,84 @@ def test_abort(self):
121 122
         ar2.abort()
122 123
         self.assertRaises(error.TaskAborted, ar2.get)
123 124
         ar.get()
  125
+    
  126
+    def test_len(self):
  127
+        v = self.client.load_balanced_view()
  128
+        ar = v.map_async(lambda x: x, range(10))
  129
+        self.assertEquals(len(ar), 10)
  130
+        ar = v.apply_async(lambda x: x, range(10))
  131
+        self.assertEquals(len(ar), 1)
  132
+        ar = self.client[:].apply_async(lambda x: x, range(10))
  133
+        self.assertEquals(len(ar), len(self.client.ids))
  134
+    
  135
+    def test_wall_time_single(self):
  136
+        v = self.client.load_balanced_view()
  137
+        ar = v.apply_async(time.sleep, 0.25)
  138
+        self.assertRaises(TimeoutError, getattr, ar, 'wall_time')
  139
+        ar.get(2)
  140
+        self.assertTrue(ar.wall_time < 1.)
  141
+        self.assertTrue(ar.wall_time > 0.2)
  142
+
  143
+    def test_wall_time_multi(self):
  144
+        self.minimum_engines(4)
  145
+        v = self.client[:]
  146
+        ar = v.apply_async(time.sleep, 0.25)
  147
+        self.assertRaises(TimeoutError, getattr, ar, 'wall_time')
  148
+        ar.get(2)
  149
+        self.assertTrue(ar.wall_time < 1.)
  150
+        self.assertTrue(ar.wall_time > 0.2)
  151
+
  152
+    def test_serial_time_single(self):
  153
+        v = self.client.load_balanced_view()
  154
+        ar = v.apply_async(time.sleep, 0.25)
  155
+        self.assertRaises(TimeoutError, getattr, ar, 'serial_time')
  156
+        ar.get(2)
  157
+        self.assertTrue(ar.serial_time < 0.5)
  158
+        self.assertTrue(ar.serial_time > 0.2)
  159
+
  160
+    def test_serial_time_multi(self):
  161
+        self.minimum_engines(4)
  162
+        v = self.client[:]
  163
+        ar = v.apply_async(time.sleep, 0.25)
  164
+        self.assertRaises(TimeoutError, getattr, ar, 'serial_time')
  165
+        ar.get(2)
  166
+        self.assertTrue(ar.serial_time < 2.)
  167
+        self.assertTrue(ar.serial_time > 0.8)
  168
+
  169
+    def test_elapsed_single(self):
  170
+        v = self.client.load_balanced_view()
  171
+        ar = v.apply_async(time.sleep, 0.25)
  172
+        while not ar.ready():
  173
+            time.sleep(0.01)
  174
+            self.assertTrue(ar.elapsed < 0.3)
  175
+        self.assertTrue(ar.elapsed < 0.3)
  176
+        ar.get(2)
  177
+
  178
+    def test_elapsed_multi(self):
  179
+        v = self.client[:]
  180
+        ar = v.apply_async(time.sleep, 0.25)
  181
+        while not ar.ready():
  182
+            time.sleep(0.01)
  183
+            self.assertTrue(ar.elapsed < 0.3)
  184
+        self.assertTrue(ar.elapsed < 0.3)
  185
+        ar.get(2)
  186
+
  187
+    def test_hubresult_timestamps(self):
  188
+        self.minimum_engines(4)
  189
+        v = self.client[:]
  190
+        ar = v.apply_async(time.sleep, 0.25)
  191
+        ar.get(2)
  192
+        rc2 = Client(profile='iptest')
  193
+        # must have try/finally to close second Client, otherwise
  194
+        # will have dangling sockets causing problems
  195
+        try:
  196
+            time.sleep(0.25)
  197
+            hr = rc2.get_result(ar.msg_ids)
  198
+            self.assertTrue(hr.elapsed > 0., "got bad elapsed: %s" % hr.elapsed)
  199
+            hr.get(1)
  200
+            self.assertTrue(hr.wall_time < ar.wall_time + 0.2, "got bad wall_time: %s > %s" % (hr.wall_time, ar.wall_time))
  201
+            self.assertEquals(hr.serial_time, ar.serial_time)
  202
+        finally:
  203
+            rc2.close()
124 204
 
125 205
 
18  IPython/parallel/tests/test_client.py
@@ -316,4 +316,22 @@ def test_purge_all_results(self):
316 316
         self.client.purge_results('all')
317 317
         hist = self.client.hub_history()
318 318
         self.assertEquals(len(hist), 0)
  319
+    
  320
+    def test_spin_thread(self):
  321
+        self.client.spin_thread(0.01)
  322
+        ar = self.client[-1].apply_async(lambda : 1)
  323
+        time.sleep(0.1)
  324
+        self.assertTrue(ar.wall_time < 0.1,
  325
+            "spin should have kept wall_time < 0.1, but got %f" % ar.wall_time
  326
+        )
  327
+    
  328
+    def test_stop_spin_thread(self):
  329
+        self.client.spin_thread(0.01)
  330
+        self.client.stop_spin_thread()
  331
+        ar = self.client[-1].apply_async(lambda : 1)
  332
+        time.sleep(0.15)
  333
+        self.assertTrue(ar.wall_time > 0.1,
  334
+            "Shouldn't be spinning, but got wall_time=%f" % ar.wall_time
  335
+        )
  336
+    
319 337
 
274  docs/examples/notebooks/Animations_and_Progress.ipynb
... ...
@@ -0,0 +1,274 @@
  1
+{
  2
+ "metadata": {
  3
+  "name": "Animations_and_Progress"
  4
+ },
  5
+ "nbformat": 3,
  6
+ "worksheets": [
  7
+  {
  8
+   "cells": [
  9
+    {
  10
+     "cell_type": "heading",
  11
+     "level": 1,
  12
+     "source": [
  13
+      "Simple animations, progress bars, and clearing output"
  14
+     ]
  15
+    },
  16
+    {
  17
+     "cell_type": "markdown",
  18
+     "source": [
  19
+      "Sometimes you want to print progress in-place, but don't want",
  20
+      "to keep growing the output area.  In terminals, there is the carriage-return",
  21
+      "(`'\\r'`) for overwriting a single line, but the notebook frontend does not support this",
  22
+      "behavior (yet).",
  23
+      "",
  24
+      "What the notebook *does* support is explicit `clear_output`, and you can use this to replace previous",
  25
+      "output (specifying stdout/stderr or the special IPython display outputs)."
  26
+     ]
  27
+    },
  28
+    {
  29
+     "cell_type": "markdown",
  30
+     "source": [
  31
+      "A simple example printing our progress iterating through a list:"
  32
+     ]
  33
+    },
  34
+    {
  35
+     "cell_type": "code",
  36
+     "collapsed": true,
  37
+     "input": [
  38
+      "import sys",
  39
+      "import time"
  40
+     ],
  41
+     "language": "python",
  42
+     "outputs": [],
  43
+     "prompt_number": 16
  44
+    },
  45
+    {
  46
+     "cell_type": "code",
  47
+     "collapsed": false,
  48
+     "input": [
  49
+      "from IPython.core.display import clear_output",
  50
+      "for i in range(10):",
  51
+      "    time.sleep(0.25)",
  52
+      "    clear_output()",
  53
+      "    print i",
  54
+      "    sys.stdout.flush()"
  55
+     ],
  56
+     "language": "python",
  57
+     "outputs": [],
  58
+     "prompt_number": 12
  59
+    },
  60
+    {
  61
+     "cell_type": "markdown",
  62
+     "source": [
  63
+      "The AsyncResult object has a special `wait_interactive()` method, which prints its progress interactively,",
  64
+      "so you can watch as your parallel computation completes.",
  65
+      "",
  66
+      "**This example assumes you have an IPython cluster running, which you can start from the [cluster panel](/#tab2)**"
  67
+     ]
  68
+    },
  69
+    {
  70
+     "cell_type": "code",
  71
+     "collapsed": false,
  72
+     "input": [
  73
+      "from IPython import parallel",
  74
+      "rc = parallel.Client()",
  75
+      "view = rc.load_balanced_view()",
  76
+      "",
  77
+      "amr = view.map_async(time.sleep, [0.5]*100)",
  78
+      "",
  79
+      "amr.wait_interactive()"
  80
+     ],
  81
+     "language": "python",
  82
+     "outputs": [],
  83
+     "prompt_number": 13
  84
+    },
  85
+    {
  86
+     "cell_type": "markdown",
  87
+     "source": [
  88
+      "You can also use `clear_output()` to clear figures and plots.",
  89
+      "",
  90
+      "This time, we need to make sure we are using inline pylab (**requires matplotlib**)"
  91
+     ]
  92
+    },
  93
+    {
  94
+     "cell_type": "code",
  95
+     "collapsed": false,
  96
+     "input": [
  97
+      "%pylab inline"
  98
+     ],
  99
+     "language": "python",
  100
+     "outputs": [
  101
+      {
  102
+       "output_type": "stream",
  103
+       "stream": "stdout",
  104
+       "text": [
  105
+        "",
  106
+        "Welcome to pylab, a matplotlib-based Python environment [backend: module://IPython.zmq.pylab.backend_inline].",
  107
+        "For more information, type 'help(pylab)'."
  108
+       ]
  109
+      }
  110
+     ],
  111
+     "prompt_number": 17
  112
+    },
  113
+    {
  114
+     "cell_type": "code",
  115
+     "collapsed": false,
  116
+     "input": [
  117
+      "from scipy.special import jn",
  118
+      "x = linspace(0,5)",
  119
+      "f, ax = plt.subplots()",
  120
+      "ax.set_title(\"Bessel functions\")",
  121
+      "",
  122
+      "for n in range(1,10):",
  123
+      "    time.sleep(1)",
  124
+      "    ax.plot(x, jn(x,n))",
  125
+      "    clear_output()",
  126
+      "    display(f)",
  127
+      "",
  128
+      "# close the figure at the end, so we don't get a duplicate",
  129
+      "# of the last plot",
  130
+      "plt.close()"
  131
+     ],
  132
+     "language": "python",
  133
+     "outputs": [
  134
+      {
  135
+       "output_type": "display_data",
  136
+       "png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAEICAYAAABRSj9aAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nYII=\n",
  137
+       "text": [
  138
+        "<matplotlib.figure.Figure at 0x1083272d0>"
  139
+       ]
  140
+      }
  141
+     ],
  142
+     "prompt_number": 20
  143
+    },
  144
+    {
  145
+     "cell_type": "heading",
  146
+     "level": 2,
  147
+     "source": [
  148
+      "A Javascript Progress Bar"
  149
+     ]
  150
+    },
  151
+    {
  152
+     "cell_type": "markdown",
  153
+     "source": [
  154
+      "`clear_output()` is still something of a hack, and if you want to do a progress bar in the notebook",
  155
+      "it is better to just use Javascript/HTML if you can.",
  156
+      "",
  157
+      "Here is a simple progress bar using HTML/Javascript:"
  158
+     ]
  159
+    },
  160
+    {
  161
+     "cell_type": "code",
  162
+     "collapsed": false,
  163
+     "input": [
  164
+      "import uuid",
  165
+      "from IPython.core.display import HTML, Javascript, display",
  166
+      "",
  167
+      "divid = str(uuid.uuid4())",
  168
+      "",
  169
+      "pb = HTML(",
  170
+      "\"\"\"",
  171
+      "<div style=\"border: 1px solid black; width:500px\">",
  172
+      "  <div id=\"%s\" style=\"background-color:blue; width:0%%\">&nbsp;</div>",
  173
+      "</div> ",
  174
+      "\"\"\" % divid)",
  175
+      "display(pb)",
  176
+      "for i in range(1,101):",
  177
+      "    time.sleep(0.1)",
  178
+      "    ",
  179
+      "    display(Javascript(\"$('div#%s').width('%i%%')\" % (divid, i)))"
  180
+     ],
  181
+     "language": "python",
  182
+     "outputs": [],
  183
+     "prompt_number": 15
  184
+    },
  185
+    {
  186
+     "cell_type": "markdown",
  187
+     "source": [
  188
+      "The above simply makes a div that is a box, and a blue div inside it with a unique ID ",
  189
+      "(so that the javascript won't collide with other similar progress bars on the same page).  ",
  190
+      "",
  191
+      "Then, at every progress point, we run a simple jQuery call to resize the blue box to",
  192
+      "the appropriate fraction of the width of its containing box, and voil\u00e0 a nice",
  193
+      "HTML/Javascript progress bar!"
  194
+     ]
  195
+    },
  196
+    {
  197
+     "cell_type": "heading",
  198
+     "level": 2,
  199
+     "source": [
  200
+      "ProgressBar class"
  201
+     ]
  202
+    },
  203
+    {
  204
+     "cell_type": "markdown",
  205
+     "source": [
  206
+      "And finally, here is a progress bar *class* extracted from [PyMC](http://code.google.com/p/pymc/), which will work in regular Python as well as in the IPython Notebook"
  207
+     ]
  208
+    },
  209
+    {
  210
+     "cell_type": "code",
  211
+     "collapsed": true,
  212
+     "input": [
  213
+      "import sys, time",
  214
+      "try:",
  215
+      "    from IPython.core.display import clear_output",
  216
+      "    have_ipython = True",
  217
+      "except ImportError:",
  218
+      "    have_ipython = False",
  219
+      "",
  220
+      "class ProgressBar:",
  221
+      "    def __init__(self, iterations):",
  222
+      "        self.iterations = iterations",
  223
+      "        self.prog_bar = '[]'",
  224
+      "        self.fill_char = '*'",
  225
+      "        self.width = 40",
  226
+      "        self.__update_amount(0)",
  227
+      "        if have_ipython:",
  228
+      "            self.animate = self.animate_ipython",
  229
+      "        else:",
  230
+      "            self.animate = self.animate_noipython",
  231
+      "",
  232
+      "    def animate_ipython(self, iter):",
  233
+      "        clear_output()",
  234
+      "        print '\\r', self,",
  235
+      "        sys.stdout.flush()",
  236
+      "        self.update_iteration(iter + 1)",
  237
+      "",
  238
+      "    def update_iteration(self, elapsed_iter):",
  239
+      "        self.__update_amount((elapsed_iter / float(self.iterations)) * 100.0)",
  240
+      "        self.prog_bar += '  %d of %s complete' % (elapsed_iter, self.iterations)",
  241
+      "",
  242
+      "    def __update_amount(self, new_amount):",
  243
+      "        percent_done = int(round((new_amount / 100.0) * 100.0))",
  244
+      "        all_full = self.width - 2",
  245
+      "        num_hashes = int(round((percent_done / 100.0) * all_full))",
  246
+      "        self.prog_bar = '[' + self.fill_char * num_hashes + ' ' * (all_full - num_hashes) + ']'",
  247
+      "        pct_place = (len(self.prog_bar) / 2) - len(str(percent_done))",
  248
+      "        pct_string = '%d%%' % percent_done",
  249
+      "        self.prog_bar = self.prog_bar[0:pct_place] + \\",
  250
+      "            (pct_string + self.prog_bar[pct_place + len(pct_string):])",
  251
+      "",
  252
+      "    def __str__(self):",
  253
+      "        return str(self.prog_bar)"
  254
+     ],
  255
+     "language": "python",
  256
+     "outputs": [],
  257
+     "prompt_number": 10
  258
+    },
  259
+    {
  260
+     "cell_type": "code",
  261
+     "collapsed": false,
  262
+     "input": [
  263
+      "p = ProgressBar(1000)",
  264
+      "for i in range(1001):",
  265
+      "    p.animate(i)"
  266
+     ],
  267
+     "language": "python",
  268
+     "outputs": [],
  269
+     "prompt_number": 11
  270
+    }
  271
+   ]
  272
+  }
  273
+ ]
  274
+}
116  docs/source/parallel/asyncresult.txt
... ...
@@ -0,0 +1,116 @@
  1
+.. _parallel_asyncresult:
  2
+
  3
+======================
  4
+The AsyncResult object
  5
+======================
  6
+
  7
+In non-blocking mode, :meth:`apply` submits the command to be executed and
  8
+then returns a :class:`~.AsyncResult` object immediately. The
  9
+AsyncResult object gives you a way of getting a result at a later
  10
+time through its :meth:`get` method, but it also collects metadata
  11
+on execution.
  12
+
  13
+
  14
+Beyond multiprocessing's AsyncResult
  15
+====================================
  16
+
  17
+.. Note::
  18
+
  19
+    The :class:`~.AsyncResult` object provides a superset of the interface in 
  20
+    :py:class:`multiprocessing.pool.AsyncResult`.  See the 
  21
+    `official Python documentation <http://docs.python.org/library/multiprocessing#multiprocessing.pool.AsyncResult>`_
  22
+    for more on the basics of this interface.
  23
+
  24
+Our AsyncResult objects add a number of convenient features for working with
  25
+parallel results, beyond what is provided by the original AsyncResult.
  26
+
  27
+
  28
+get_dict
  29
+--------
  30
+
  31
+First, is :meth:`.AsyncResult.get_dict`, which pulls results as a dictionary
  32
+keyed by engine_id, rather than a flat list.  This is useful for quickly
  33
+coordinating or distributing information about all of the engines.
  34
+
  35
+As an example, here is a quick call that gives every engine a dict showing
  36
+the PID of every other engine:
  37
+
  38
+.. sourcecode:: ipython
  39
+
  40
+    In [10]: ar = rc[:].apply_async(os.getpid)
  41
+    In [11]: pids = ar.get_dict()
  42
+    In [12]: rc[:]['pid_map'] = pids
  43
+
  44
+This trick is particularly useful when setting up inter-engine communication,
  45
+as in IPython's :file:`examples/parallel/interengine` examples.
  46
+
  47
+
  48
+Metadata
  49
+========
  50
+
  51
+IPython.parallel tracks some metadata about the tasks, which is stored
  52
+in the :attr:`.Client.metadata` dict.  The AsyncResult object gives you an
  53
+interface for this information as well, including timestamps stdout/err,
  54
+and engine IDs.
  55
+
  56
+
  57
+Timing
  58
+------
  59
+
  60
+IPython tracks various timestamps as :py:class:`.datetime` objects,
  61
+and the AsyncResult object has a few properties that turn these into useful
  62
+times (in seconds as floats).
  63
+
  64
+For use while the tasks are still pending:
  65
+
  66
+* :attr:`ar.elapsed` is just the elapsed seconds since submission, for use
  67
+  before the AsyncResult is complete.
  68
+* :attr:`ar.progress` is the number of tasks that have completed.  Fractional progress
  69
+  would be::
  70
+
  71
+    1.0 * ar.progress / len(ar)
  72
+
  73
+* :meth:`AsyncResult.wait_interactive` will wait for the result to finish, but
  74
+  print out status updates on progress and elapsed time while it waits.
  75
+
  76
+For use after the tasks are done:
  77
+
  78
+* :attr:`ar.serial_time` is the sum of the computation time of all of the tasks
  79
+  done in parallel.
  80
+* :attr:`ar.wall_time` is the time between the first task submitted and last result
  81
+  received.  This is the actual cost of computation, including IPython overhead. 
  82
+  
  83
+
  84
+.. note::
  85
+
  86
+  wall_time is only precise if the Client is waiting for results when
  87
+  the task finished, because the `received` timestamp is made when the result is
  88
+  unpacked by the Client, triggered by the :meth:`~Client.spin` call. If you
  89
+  are doing work in the Client, and not waiting/spinning, then `received` might
  90
+  be artificially high.
  91
+
  92
+An often interesting metric is the time it actually cost to do the work in parallel
  93
+relative to the serial computation, and this can be given simply with
  94
+
  95
+.. sourcecode:: python
  96
+
  97
+    speedup = ar.serial_time / ar.wall_time
  98
+
  99
+
  100
+Map results are iterable!
  101
+=========================
  102
+
  103
+When an AsyncResult object has multiple results (e.g. the :class:`~AsyncMapResult`
  104
+object), you can actually iterate through them, and act on the results as they arrive:
  105
+
  106
+.. literalinclude:: ../../examples/parallel/itermapresult.py
  107
+    :language: python
  108
+    :lines: 20-66
  109
+
  110
+.. seealso::
  111
+
  112
+    When AsyncResult or the AsyncMapResult don't provide what you need (for instance,
  113
+    handling individual results as they arrive, but with metadata), you can always
  114
+    just split the original result's ``msg_ids`` attribute, and handle them as you like.
  115
+    
  116
+    For an example of this, see :file:`docs/examples/parallel/customresult.py`
1  docs/source/parallel/index.txt
@@ -11,6 +11,7 @@ Using IPython for parallel computing
11 11
    parallel_process.txt
12 12
    parallel_multiengine.txt
13 13
    parallel_task.txt
  14
+   asyncresult.txt
14 15
    parallel_mpi.txt
15 16
    parallel_db.txt
16 17
    parallel_security.txt
8  docs/source/parallel/parallel_multiengine.txt
@@ -275,13 +275,9 @@ then returns a :class:`AsyncResult` object immediately. The
275 275
 :class:`AsyncResult` object gives you a way of getting a result at a later
276 276
 time through its :meth:`get` method.
277 277
 
278  
-.. Note::
279  
-
280  
-    The :class:`AsyncResult` object provides a superset of the interface in 
281  
-    :py:class:`multiprocessing.pool.AsyncResult`.  See the 
282