- Pharo-dev - lists.pharo.org

Re: [Pharo-dev] Peek

by StÃ©phane Ducasse

> > > BTW, do we still need UTF16 support ? > > For those encodings that we still want to support in the future, we should have new and more principled implementations under ZnCharacterEncoder. That is, if we ever want to fase out TextConverter. :) You know what I think. Kill all the rotten code. Stef

Nov. 6, 2013

Re: [Pharo-dev] Peek

by StÃ©phane Ducasse

Hi henrik why don't you give a try to change our lives and propose a new MultuByteFileStream and friends :) Stef > That's great! > Remembering that commit message was part of the reason for benching, was sort of disappointed there was no significant difference between Zn in 2.0 and latest 3.0... > > I guess with the amount of hacks accumulating, it is indeed turning into a worthy successor of MultiByteFileStream ;) > > Cheers, > Henry > > P.S: If you want another delightful one in the same vein (both from WTFy-ness and perf improvement POV), take a gander at UTF16TextConverter >> nextPutByteString:toStream: > > > On Mon, Nov 4, 2013 at 5:12 PM, Sven Van Caekenberghe <sven(a)stfx.eu> wrote: > Hi Henrik, > > Great writeup, thanks ! > > (more inline) > > On 04 Nov 2013, at 11:58, Henrik Johansen <henrik.s.johansen(a)veloxit.no> wrote: > > > On 04 Nov 2013, at 9:57 , Diego Lont <diego.lont(a)delware.nl> wrote: > > > >> Working on Petit Delphi we found a strange implementation for asPetitStream: > >> Stream>asPetitStream > >> ^ self contents asPetitStream > >> > >> Further investigation showed that the basic peek was not fast enough for Petit Parser, as it is used a lot. So it implemented a "improved unchecked peek": > >> PPStream>peek > >> "An improved version of peek, that is slightly faster than the built in version." > >> ^ self atEnd ifFalse: [ collection at: position + 1 ] > >> > >> PPStream>uncheckedPeek > >> "An unchecked version of peek that throws an error if we try to peek over the end of the stream, even faster than #peek." > >> ^ collection at: position + 1 > >> > >> But in my knowledge a basic peek should be fast. The real problem is the peek in the underlying peek: > >> PositionableStream>peek > >> "Answer what would be returned if the message next were sent to the > >> receiver. If the receiver is at the end, answer nil." > >> > >> | nextObject | > >> self atEnd ifTrue: [^nil]. > >> nextObject := self next. > >> position := position - 1. > >> ^nextObject > >> > >> That actually uses "self next". The least thing one should do is to cache the next object. But isn't there a primitive for peek in a file stream? Because al overriding peeks of PositionableStream have basically the same implementation: reading the next and restoring the state to before the peek (that is slow). So we would like to be able to remove PPStream without causing performance issues, as the only added method is the "improved peek". > >> > >> Stephan and Diego > > > > If you are reading from file, ZnCharacterStream should be a valid alternative. > > If not, ZnBufferedReadStream on an internal collection stream also does peek caching. > > > > Beware with files though; itâs better to bench the overall operation for different alternatives. > > F.ex, ZnCharacterStream is much faster than the standard Filestream for peek: > > > > cr := ZnCharacterReadStream on: 'PharoDebug.log' asFileReference readStream binary. > > [cr peek] bench. '49,400,000 per second.' > > cr close. > > > > FileStream fileNamed: 'PharoDebug.log' do: [:fs | [fs peek] bench] '535,000 per second.â > > > > but has different bulk reads characteristics (faster for small bulks, slower for large bulks, crossover-point at around 1k chars at once); > > (The actual values are of course also dependent on encoder/file contents, those given here obtained with UTF-8 and a mostly/all ascii text file) > > > > [cr := ZnCharacterReadStream on: ('PharoDebug.log' asFileReference readStream binary ) readStream. > > cr next: 65536; close] bench '105 per second.' '106 per second.â > > Well, I just realised that ZnCharacterReadStream and ZnCharacterWriteStream did not yet make use of the optimisations that I did for ZnCharacterEncoding some time ago. More specifically, they were not yet using #next:putAll:startingAt:toStream: and #readInto:startingAt:count:fromStream: which are overwritten for ZnUTF8Encoder with (super hacked) versions that assume most of the input will be ASCII (a reasonable assumption). > > I am still chasing a bug, but right now: > > [ (ZnCharacterReadStream on: ('timezones.json' asFileReference readStream binary)) > next: 65536; close ] bench. > > "135 per second.â BEFORE > "3,310 per second.â AFTER > > But of course the input file is ASCII, so YMMV. > > Iâll let you know when I commit this code. > > Sven > > > [FileStream fileNamed: 'PharoDebug.log' do: [:fs | fs next: 65536] > > ] bench '176 per second.â > > > > If you use a StandardFilestream set to binary ( which has less overhead for binary nextâs compared to the MultiByteFileStream returned by asFileReference readStream)as the base stream instead, , the same general profile holds true, but with a crossover around 2k characters. > > > > TL;DR: Benchmark the alternatives. The best replacement option depends on your results. Appropriately (according to source and actual use) set up Zn-streams are probably your best bet. > > > > Cheers, > > Henry > > >

Nov. 6, 2013

Re: [Pharo-dev] Support understanding changes

by StÃ©phane Ducasse

On Nov 6, 2013, at 1:38 PM, Goubier Thierry <thierry.goubier(a)cea.fr> wrote: > > > Le 06/11/2013 13:20, StÃ©phane Ducasse a Ã©crit : >> Now that we have epicea I would really love to have a tool that does not show me stupidly a diff but >> take into account the actions that have been performed like rename class, splitâ¦. > > Yes! I want that too! How do we try EPICEA? Is it already integrated? Martin did not send a public mail to the list? Shame on him :) >> Side questions: camillo and other giter, I was thinking that it would make sense to publish on git metadata (may be in ston) >> representing the semantics of the operation that led to the changes) so that tools can take advantage of this information >> to present semantical operation instead of plain stupid diff. For example split this method, rename class,â¦. > > +100 > > My approach would be to focus on tools in the Pharo world to explore that, at the Monticello GUI level (and merge tools). Yes you are right. But people now use git to browse changes and it only work for little edition. But you are right. I would like also to have that combined with Torch because the Torch visualziation was so great. > In the Git world, I'll focus simply on a file format able to store that knowledge, in a way which minimises git-induced conflicts... The goal being that a git merge would recreate a working Smalltalk result (no conflicts) with a correct EPICEA history (eventually recreated from the git stored data). > > What I know is that a single file in whatever format containing the EPICEA log will be a conflict magnet. Why conflicting we will change it systematically. > > Thierry > -- > Thierry Goubier > CEA list > Laboratoire des Fondations des SystÃ¨mes Temps RÃ©el EmbarquÃ©s > 91191 Gif sur Yvette Cedex > France > Phone/Fax: +33 (0) 1 69 08 32 92 / 83 95 >

Nov. 6, 2013

Re: [Pharo-dev] Support understanding changes

by Goubier Thierry

Le 06/11/2013 13:20, StÃ©phane Ducasse a Ã©crit : > Now that we have epicea I would really love to have a tool that does not show me stupidly a diff but > take into account the actions that have been performed like rename class, splitâ¦. Yes! I want that too! How do we try EPICEA? Is it already integrated? > Side questions: camillo and other giter, I was thinking that it would make sense to publish on git metadata (may be in ston) > representing the semantics of the operation that led to the changes) so that tools can take advantage of this information > to present semantical operation instead of plain stupid diff. For example split this method, rename class,â¦. +100 My approach would be to focus on tools in the Pharo world to explore that, at the Monticello GUI level (and merge tools). In the Git world, I'll focus simply on a file format able to store that knowledge, in a way which minimises git-induced conflicts... The goal being that a git merge would recreate a working Smalltalk result (no conflicts) with a correct EPICEA history (eventually recreated from the git stored data). What I know is that a single file in whatever format containing the EPICEA log will be a conflict magnet. Thierry -- Thierry Goubier CEA list Laboratoire des Fondations des SystÃ¨mes Temps RÃ©el EmbarquÃ©s 91191 Gif sur Yvette Cedex France Phone/Fax: +33 (0) 1 69 08 32 92 / 83 95

Nov. 6, 2013

Re: [Pharo-dev] Peek

by Henrik Johansen

On 06 Nov 2013, at 10:38 , Sven Van Caekenberghe <sven(a)stfx.eu> wrote: > BTW, do we still need UTF16 support ? > > For those encodings that we still want to support in the future, we should have new and more principled implementations under ZnCharacterEncoder. That is, if we ever want to fase out TextConverter. UTF16 is the encoding for string arguments to the Unicode-aware version of the Windows API's, so Iâd say yes. Not too sure about JNI, but I know Java also uses UTF16 internally. So the short answer is yes. Not as a storage encoding, but for calling foreign functions, itâs quite important. Cheers, Henry

Nov. 6, 2013

Re: [Pharo-dev] Support understanding changes

by StÃ©phane Ducasse

I think that we could also add such metadata in MCZ files and it would be great. This is an excellent topics for an intern. Stef On Nov 6, 2013, at 1:20 PM, StÃ©phane Ducasse <stephane.ducasse(a)inria.fr> wrote: > Now that we have epicea I would really love to have a tool that does not show me stupidly a diff but > take into account the actions that have been performed like rename class, splitâ¦. > > Side questions: camillo and other giter, I was thinking that it would make sense to publish on git metadata (may be in ston) > representing the semantics of the operation that led to the changes) so that tools can take advantage of this information > to present semantical operation instead of plain stupid diff. For example split this method, rename class,â¦. > > Stef

Nov. 6, 2013

Re: [Pharo-dev] Deep Into Pharo: PPExpressionParser

by StÃ©phane Ducasse

On Nov 5, 2013, at 10:11 PM, Sean P. DeNigris <sean(a)clipperadams.com> wrote: > After reading the PetitParser chapter, I'm not clear when one would subclass > PPCompositeParser vs. using PPExpressionParser. What are the tradeoffs / use > cases? Also, can the book be improved or is that the final version? Thanks. The book is to be improved please help :) > > > > ----- > Cheers, > Sean > -- > View this message in context: http://forum.world.st/Deep-Into-Pharo-PPExpressionParser-tp4719454.html > Sent from the Pharo Smalltalk Developers mailing list archive at Nabble.com. >

Nov. 6, 2013

Support understanding changes

by StÃ©phane Ducasse

Now that we have epicea I would really love to have a tool that does not show me stupidly a diff but take into account the actions that have been performed like rename class, splitâ¦. Side questions: camillo and other giter, I was thinking that it would make sense to publish on git metadata (may be in ston) representing the semantics of the operation that led to the changes) so that tools can take advantage of this information to present semantical operation instead of plain stupid diff. For example split this method, rename class,â¦. Stef

Nov. 6, 2013

Re: [Pharo-dev] Peek

by Henrik Johansen

On 04 Nov 2013, at 5:12 , Sven Van Caekenberghe <sven(a)stfx.eu> wrote: > > Well, I just realised that ZnCharacterReadStream and ZnCharacterWriteStream did not yet make use of the optimisations that I did for ZnCharacterEncoding some time ago. More specifically, they were not yet using #next:putAll:startingAt:toStream: and #readInto:startingAt:count:fromStream: which are overwritten for ZnUTF8Encoder with (super hacked) versions that assume most of the input will be ASCII (a reasonable assumption). > > I am still chasing a bug, but right now: > > [ (ZnCharacterReadStream on: ('timezones.json' asFileReference readStream binary)) > next: 65536; close ] bench. > > "135 per second.â BEFORE > "3,310 per second.â AFTER > > But of course the input file is ASCII, so YMMV. > > Iâll let you know when I commit this code. > > Sven Yeahâ¦ sooo, I loaded the updated version, great improvement for streams on Latin1 content :D Maybe itâs just me, but I tested with actual wide source as well (it was as slow as youâd expect), and I think you need a notQuiteSoOptimizedReadInto* which uses the normal Byte -> Wide become: conversion machinery. Writing a ZnByteStringBecameWideString handler for every next: (and cousins) call where source may be non-latin1 is a real chore/nasty surprise for those used to dealing with legacy streams/converters... You can sort of kinda make up for the performance hit (at least on these sizes) with a faster replace:from:to:with:startingAt: in use after the fact, by using basicAt:put: to the string, thus avoiding converting the replacement value to character. Cheers, Henry PS: Why is ZnByteStringBecameWideString a notification and not a resumable exception? I would assume those who run into it without a handler would rather have an actual error, than a result where their string has been read up to the first wide char...

Nov. 6, 2013

Re: [Pharo-dev] No decompile string at all, bad idea?

by Mariano Martinez Peck

On Wed, Nov 6, 2013 at 4:57 AM, Marcus Denker <marcus.denker(a)inria.fr>wrote: > > On 06 Nov 2013, at 08:53, Max Leske <maxleske(a)gmail.com> wrote: > > > > > On 06.11.2013, at 08:05, Marcus Denker <marcus.denker(a)inria.fr> wrote: > > > >> > >> On 05 Nov 2013, at 20:37, Mariano Martinez Peck <marianopeck(a)gmail.com> > wrote: > >> > >>> Hi. From what I understand you removed the old Compiler but yet Opal > does not support compiling. Also, #sourceCode was changed etc... > >>> That means that there is no way to see the decompiled string of a > method? > >> > >> Yes. > >> > >>> Which means I cannot deploy and image without sources and then > browse/debug/inspect/write pharoDebug.log etc because we don't have even > the decompile string??? > >> > >> Exactly. But this is just temporary, .sources and .changes will go away > in Pharo4 and there will be just a .pharo4 image file. > > > > By âtemporaryâ do you mean decompilation will come back? > > > No, we will have a high-level representation of Methods that replace (from > the standpoint of reflection) the incomplete bytecode representation, > > OK. This is the key from my point of view. I have always seen methods like 2 things: 1) the runtime execution part of a method (you have bytecodes, you can evaluate it, etc etc) 2) the method as source code (which has the temps also and accepts all types of things regarding source code management). So I am all in of having that high-level representation of methods, however, I am not sure if I would like to loose a representation of methods that just means runtime/bytecodes. > >> > >>> Serializing compiled methods is not fun either since we cannot > decompile it after... > >>> > >> > >> You can embedd the source into the method before serialising. > > > > That means storing redundant information though (if the user is only > interested in functionality) and that will increase file size, and > serialization/materalzation time. So itâs not an ideal situation. > > The thing is that we can either not move or move step by step taking > (sometimes) into account that things are sub-optimal for a little while. > > You can envision the current implementation as a peak on a map: itâs a > very good local optimum. But *Much* better is possible. Maybne a local > metaphor: We are in the Gurten now, you can > not reach the Eiger without going down for a while. > > Marcus > -- Mariano http://marianopeck.wordpress.com

Nov. 6, 2013