Hi��Thierry,

On Mon, Feb 27, 2017 at 12:15 AM, Thierry Goubier <thierry.goubier@gmail.com> wrote:

Hi Andrei,

Le 26/02/2017 �� 23:48, Andrei Chis a ��crit :

Hi Thierry,

Thanks for the measurements. Things are more clear now :)

On Fri, Feb 24, 2017 at 11:52 PM, Thierry Goubier
<thierry.goubier@gmail.com <mailto:thierry.goubier@gmail.com>> wrote:

�� Hi Andrei,

�� Le 24/02/2017 �� 17:02, Andrei Chis a ��crit :

�� Hi Thierry,

�� Strangely enough I'm getting different times on my machine
�� Just to start from the same baseline in a fresh Pharo 60411
�� image with
�� no changes to any of the inspectors over a series of runs I get:

�� array := (1 to: 1000000) asArray.
�� [array inspect] timeToRun.�� "0:00:00:00.145"
�� [GTInspector inspect: array] timeToRun. "0:00:00:00.031"

�� Yes, this is expected. I'm not comparing to the standard
�� EyeInspector, but with my experiments with a FastTable-derived
�� widget and the EyeInspector model. GTInspector is a lot faster than
�� the normal EyeInspector.

�� As default (the only one displaying 100k elements is the AltInspector)

�� | array |
�� array := (1 to: 1000000) asArray.
�� [AltInspector inspect: array ] timeToRun. "0:00:00:00.096"
�� [EyeInspector inspect: array] timeToRun.�� "0:00:00:00.364"
�� [GTInspector inspect: array] timeToRun. "0:00:00:00.065"

Indeed this is what I expected.

�� If I change indexableDisplayLimit to 50000 and remove the Items view
�� then I get:

�� [GTInspector inspect: array] timeToRun.�� "0:00:00:00.124"

�� [GTInspector inspect: array] timeToRun. "0:00:00:00.256"

�� So this time is to be compared to the "0:00:00:00.096".

Can you just also try to remove the Items view and set
indexableDisplayLimit to 50.

In the previous times, the Item view was removed. If indexableDisplayLimit is set at 50, then the time become:

[GTInspector inspect: array] timeToRun. "0:00:00:00.156"

Which is interesting in itself: multiplying the limit by 1000 (from 50 to 50k) increase the time by a factor of ~1.6.

You're right, the performance of the datasource it's actually quite good.

��

I think there is some small performance degradation in the current tree
data source used by Glamour but I'm not 100% sure.

It looks pretty efficient to me. There isn't much increase in times.

I looked at the scaling in my experiment: the lower bound in opening the inspector is reached for a 100k elements array. Anything smaller
has the same time (~80ms). The 1000k array increase that by ~15/20ms.

I still think ~150ms to just open the Raw view is a bit too much, but for the moment it should be fine.��

��

�� I'm really interested in seeing what makes GTInspector slower on
�� your
�� machine. Did you do other optimizations to SpecInspector?

�� The optimisations are:
�� - remove Spec (revert to a pure morphic application),
�� - use a FastTable-derived widget for a tree view (with a 100k
�� element limit),
�� - have the GT views as subitems in the widget.

Nice. Would be also interesting to see if in the future we can merge
more the models used by GTInspector and EyeInspector.

Could be. I just used them out of convenience, and I needed an inspector with the ability to view large trees.

For now, for the Raw view I set the limit to 5000 elements. This seems
fast/large enough and it allows to add the automatic refresh back to the
Raw view.

I wonder if it wouldn't be possible to tune the automatic refresh. Increase the time interval if it takes to long (for example when refreshing a set with 1M elements). But 5000 elements is already a good limit.

Tuning the refresh rate could definitely be a possibility. For now I set it to 1.5 seconds.

Cheers,

Andrei

��

Regards,

Thierry

Cheers,
Andrei

�� I'm using a MacBook Pro Retina - 2.8 GHz Intel Core i7

�� Back on a Acer Chromebook C720p 1.4GHz Intel Celeron, with vmLatest.

�� 5.0-201702231204�� Thu Feb 23 12:36:20 UTC 2017 gcc 4.6.3 [Production
�� Spur ITHB VM]
�� CoInterpreter * VMMaker.oscog-EstebanLorenzano.2136 uuid:
�� 40534c32-ca6b-4e97-91ec-31d509e49b0c Feb 23 2017
�� StackToRegisterMappingCogit * VMMaker.oscog-EstebanLorenzano.2136
�� uuid: 40534c32-ca6b-4e97-91ec-31d509e49b0c Feb 23 2017
�� VM: 201702231204 https://github.com/pharo-project/pharo-vm.git
�� <https://github.com/pharo-project/pharo-vm.git> $ Date: Thu Feb 23
�� 13:04:59 2017 +0100 $
�� Plugins: 201702231204 https://github.com/pharo-project/pharo-vm.git
�� <https://github.com/pharo-project/pharo-vm.git> $
�� Linux testing-docker-b6b0368d-4817-4638-86be-f022b8a58580
�� 4.8.12-040812-generic #201612020431 SMP Fri Dec 2 09:33:31 UTC 2016
�� i686 i686 i386 GNU/Linux

�� Pharo VM version:
�� CoInterpreter VMMaker.oscog-HolgerHansPeterFreyther.1880 uuid:
�� 16138eb3-2390-40f5-a6c8-15f0494936f8 Oct 10 2016
�� StackToRegisterMappingCogit
�� VMMaker.oscog-HolgerHansPeterFreyther.1880
�� uuid: 16138eb3-2390-40f5-a6c8-15f0494936f8 Oct 10 2016
�� git@github.com:pharo-project/pharo-vm.git Commit:
�� 06744effac0f0aa3b4b32e17636448f9d51d6707 Date: 2016-09-30
�� 08:40:43 +0200
�� By: GitHub <noreply@github.com <mailto:noreply@github.com>
�� <mailto:noreply@github.com <mailto:noreply@github.com>>>

�� Cheers,
�� Andrei

�� On Fri, Feb 24, 2017 at 3:24 PM, Thierry Goubier
�� <thierry.goubier@gmail.com <mailto:thierry.goubier@gmail.com>
�� <mailto:thierry.goubier@gmail.com
�� <mailto:thierry.goubier@gmail.com>>> wrote:

�� 2017-02-24 14:29 GMT+01:00 Andrei Chis
�� <chisvasileandrei@gmail.com <mailto:chisvasileandrei@gmail.com>
�� <mailto:chisvasileandrei@gmail.com
�� <mailto:chisvasileandrei@gmail.com>>>:

�� On Fri, Feb 24, 2017 at 12:16 PM, Thierry Goubier
�� <thierry.goubier@gmail.com
�� <mailto:thierry.goubier@gmail.com>
�� <mailto:thierry.goubier@gmail.com
�� <mailto:thierry.goubier@gmail.com>>>
�� wrote:

�� Hi Andrei,

�� 2017-02-24 11:31 GMT+01:00 Andrei Chis
�� <chisvasileandrei@gmail.com
�� <mailto:chisvasileandrei@gmail.com>
�� <mailto:chisvasileandrei@gmail.com

�� <mailto:chisvasileandrei@gmail.com>>>:

�� Hi Thierry,

�� Indeed that's the simplest option now that we
�� are using
�� fast table.

�� Just now in the case of the Raw view an
�� OrderedCollection is used to store all displayed
�� elements.
�� If you display large collections every time you
�� open the
�� Raw view it will instantiate a collection of
�� size 100k
�� and instantiate 100k objects of
�� type GTInspectorIndexedNode. With FastTable we can
�� lazily load elements so we should be able to
�� remove this
�� behaviour and the limit. Just we need to make
�� sure it
�� will play nicely with automatic refresh. There
�� is also
�� the issue that when expanding an element in the
�� tree if
�� it's a collection you don't want to expand all
�� elements.

�� I'm not sure you need to worry too much about that
�� one; in
�� practical experiments, creating that 100k collection for
�� viewing (and the associated nodes instances) isn't too
�� costly (unless creating the GTInspectorIndexedNodes has
�� hidden costs: I've only experimented with the
�� EyeInspector
�� framework).

�� There should be no hidden costs in GTInspectorIndexedNodes.
�� I made some experiments in the latest Pharo version and
�� opening
�� the Raw view on an array with one million numbers takes
�� around
�� 120ms when 100k elements are computed.
�� I'll be curious how much it takes on your machine. To test
�� update indexableDisplayLimit to 50000 in
�� Object>>#gtInspectorVariableNodesIn: and remove the
�� annotation
�� from Collection>>#gtInspectorItemsIn: (so that the Items
�� presentation is not loaded)

�� arrayLarge := (1 to: 1000000) asArray.
�� arrayLarge inspect.

�� This is the values I get on my work laptop (core i3-2350M
�� 2.30 Ghz);
�� both inspectors displays 100k elements.

�� (1 to: 1000000) asArray in: [ :s |�� [s inspect] timeToRun]
�� 0:00:00:00.064
�� (1 to: 1000000) asArray in: [:s | [GTInspector inspect: s]
�� timeToRun] 0:00:00:00.381

�� Pharo 6 60411

�� Opening the tree with all elements works fine in my
�� experiments. Tuning scrolling as done in Bloc is
�� necessary,
�� however.

�� I'm looking now on a lazy data source for
�� FastTable that
�� plays nicely with GTInspector. Let's see how it
�� will go.
�� Help is always welcomed :)

�� As I said: do not overoptimize that part... just
�� remove that
�� limitation on the raw view and measure.

�� If I measure the Items view on the previous array it takes
�� around 35ms.
�� What I'd like to have is the same opening time for the Items
�� view on large Sets and Dictionaries.

�� On my machine, the experiment is that displaying the set is
�� fast,
�� but the system becomes totally unresponsive... which may be
�� an issue
�� with the self refresh of the inspector. Yes, it was the culprit.

�� (1 to: 1000000) asSet in: [ :s |�� [s inspect] timeToRun]
�� 0:00:00:00.034
�� (1 to: 1000000) asSet in: [:s | [GTInspector inspect: s]
�� timeToRun]
�� 0:00:00:00.199

�� Regards,

�� Thierry

�� I think I used the word paginator in the wrong
�� way. If
�� you have a very large collection (millions of
�� elements)
�� I do not want to scroll through elements but
�� most likely
�� jump to a certain element or view just a subset
�� of all
�� elements. Not really add a paginator like in web
�� pages.

�� Ok, millions of elements is a bit out of my scope...
�� I'll
�� look for filtering then.

�� Yes, definitely filtering is the way to go there :)

�� Cheers,
�� Andrei

�� Regards,

�� Thierry

�� Cheers,
�� Andrei

�� On Fri, Feb 24, 2017 at 10:37 AM, Thierry Goubier
�� <thierry.goubier@gmail.com
�� <mailto:thierry.goubier@gmail.com>
�� <mailto:thierry.goubier@gmail.com
�� <mailto:thierry.goubier@gmail.com>>> wrote:

�� Hi Andrei,

�� if you're using fasttable for the Raw view, you
�� should be able to reach one 100k elements
�� without
�� issues. I did some experiments and it
�� handles the
�� load very well.

�� Avoid the paginator at any cost. This thing is
�� really user-unfriendly.

�� Regards,

�� Thierry

�� 2017-02-23 20:19 GMT+01:00 Andrei Chis
�� <chisvasileandrei@gmail.com
�� <mailto:chisvasileandrei@gmail.com>
�� <mailto:chisvasileandrei@gmail.com

�� <mailto:chisvasileandrei@gmail.com>>>:

�� Hi Stef,

�� Currently that's the default behaviour
�� of the
�� Raw view: it displays for collections
�� only the
�� first and the last 21 elements. The
�� Items view
�� however always should display all the
�� elements
�� of a collection.

�� The main problem with the Raw view in
�� Pharo 5 is
�� the speed. In Pharo 6 now the speed of
�� the Raw
�� view is greately improved so we could
�� increase
�� those limits. Still for now there should
�� still
�� be some limit for the Raw view. Ideally we
�� should add a small widget, something like a
�� paginator, for navigating through large
�� and very
�� large collections.

�� Cheers,
�� Andrei

�� On Feb 23, 2017 19:35, "stepharong"
�� <stepharong@free.fr
�� <mailto:stepharong@free.fr> <mailto:stepharong@free.fr
�� <mailto:stepharong@free.fr>>>
�� wrote:

�� Hi

�� I'm trying to debug citezen
�� generation and I
�� have to compare strings.
�� Now I think that the raw views (in
�� Pharo 50)
�� is not good because we cannot see
�� all the
�� items in raw format.
�� See the attachements. It jumps from
�� 21 to
�� 174 ...
�� and what I want to see is of course
�� in the
�� middle.

�� Is it me or there is something wrong
�� there.
�� Stef