May 2011 - Pharo-users - lists.pharo.org

Re: [Pharo-users] WebClient for 1.2.2?

by Andy Burnett

Out of curiosity, what will it take to get SSL working? Is it a difficult technical problem, or is it just a question of time/money? Cheers Andy -- View this message in context: http://forum.world.st/WebClient-for-1-2-2-tp3485946p3496034.html Sent from the Pharo Smalltalk Users mailing list archive at Nabble.com.

May 4, 2011

[ANN] PHANtom: A modern Aspect Language for Pharo

by Lemuus

Hello all, We are happy to announce the release of PHANtom: a modern aspect language for Pharo. PHANtom was built taking into account recent advances in aspect language features such as dynamic advice rescheduling, reentrancy management, and scoping strategies for deployment. The advanced features are not all implemented yet, but the 'normal' features work and are well-tested (about 130 unit tests) so we are relasing it to the world for you to try out. We hope PHANtom can help you to solve some cross-cutting code problems. An overview of the language can be found at http://pleiad.cl/PHANtom Please try it out, see if it works for you, all your comments are welcome! More details: I am doing this as my engineering thesis, under the guidance of Prof. Johan Fabry. PHANtom works in Pharo 1.2.1 and in 1.3, and also in PharoCore. Right now, PHANtom uses Method Wrappers in its' implementation, just like the AspectS language did. But the plan is once we have all features implemented, that we will start compiling using the Opal compiler of Pharo, to make everything go faster. I just put one example from the website here so you can see what it looks like: asp := PhAspect new add: (PhAdvice new pointcut: (PhPointcut new receivers: 'SystemWindow+'; selectors: 'position:'; context: #(#receiver #arguments)); advice: [:context | Transcript show: context receiver. Transcript show: ' moved to: '. Transcript show: context arguments; cr.]; type: #after). asp install. Note: the unit tests of pointcuts take a large amount of time to execute, as some of them match on everything in the image. For more info and how to download PHANtom see http://pleiad.cl/PHANtom

May 4, 2011

Re: [Pharo-users] Object>>copyTwoLevel

by andres

Hi Panu. I don't know if immutability is available in Pharo, but if it is it may work for you. You basically mark an object as immutable and register an exception handler. When an immutable object's i.v. is assigned a new value, an exception is raised and the assignment is left pending. There you can log the change and effectively apply the assignment to the object. HTH, AndrÃ©s Panu Suominen escribiÃ³: > Hi, > > we have a problem where we must create audit trail of changes made to > business objects. Also things should run inside transaction. Our first > approach was to create so called command objects from every change but > these would be hard to maintain by hand. For example there would be > ChangePersonName, ChangeXXXName classes and these would have to be > hand created in name: -mehtods. More "clever" approach we come up with > is to copy original object and use proxy that records messages send to > it and sends them to the copy. Also this proxy creates new proxy when > copy returns values. These proxies record message sends to transaction > and when transaction is ready these recorded message from all of the > proxies are done to the actual objects instead of copies. > > To create usable copies I found method that (shallow) copies the > actual object and its instance variables (Object>>copyTwoLevel). The > method is marked to be deprecated soon. > Why is that and is there some alternative way of doing this kind of copy? > Also if you know some way to trivially handle the audit trail problem > I would like to hear... :) > >

May 4, 2011

Re: [Pharo-users] Object>>copyTwoLevel

by Igor Stasenko

On 4 May 2011 15:02, Panu Suominen <panu.suominen(a)iki.fi> wrote: > Hi, > > we have a problem where we must create audit trail of changes made to > business objects. Also things should run inside transaction. Our first > approach was to create so called command objects from every change but > these would be hard to maintain by hand. For example there would be > ChangePersonName, ChangeXXXName classes and these would have to be > hand created in name: -mehtods. More "clever" approach we come up with > is to copy original object and use proxy that records messages send to > it and sends them to the copy. Also this proxy creates new proxy when > copy returns values. These proxies record message sends to transaction > and when transaction is ready these recorded message from all of the > proxies are done to the actual objects instead of copies. > > To create usable copies I found method that (shallow) copies the > actual object and its instance variables (Object>>copyTwoLevel). The > method is marked to be deprecated soon. > Why is that and is there some alternative way of doing this kind of copy? > Also if you know some way to trivially handle the audit trail problem > I would like to hear... :) > > I think that this is too special kind of copy to be provided by default in kernel (so whoever stated that this will be deprecated, i agree with him ;). What you can do is to copy down this method to the root class of your domain hierarchy.. I.e. consider that all your domain objects having single root superclass , like MyDomainObject and then you're free to use this specific kind of copy for your classes, without relying on it in Object protocol. > -- > Panu > -- Best regards, Igor Stasenko AKA sig.

May 4, 2011

Object>>copyTwoLevel

by Panu Suominen

Hi, we have a problem where we must create audit trail of changes made to business objects. Also things should run inside transaction. Our first approach was to create so called command objects from every change but these would be hard to maintain by hand. For example there would be ChangePersonName, ChangeXXXName classes and these would have to be hand created in name: -mehtods. More "clever" approach we come up with is to copy original object and use proxy that records messages send to it and sends them to the copy. Also this proxy creates new proxy when copy returns values. These proxies record message sends to transaction and when transaction is ready these recorded message from all of the proxies are done to the actual objects instead of copies. To create usable copies I found method that (shallow) copies the actual object and its instance variables (Object>>copyTwoLevel). The method is marked to be deprecated soon. Why is that and is there some alternative way of doing this kind of copy? Also if you know some way to trivially handle the audit trail problem I would like to hear... :) -- Panu

May 4, 2011

Re: [Pharo-users] About SqueakDBX

by Mariano Martinez Peck

On Wed, May 4, 2011 at 2:30 PM, Nicolas Petton <petton.nicolas(a)gmail.com>wrote: > Hi Mariano, > Hi Nicolas. Please, let me answer in the mailing list so that other can benefit also. > > I was wondering what can SqueakDBX do. http://www.squeakdbx.org/ here you have all you will probably need. In summary, it is a database driver. Like ODBC or like any other wrapper. The good thing is that it provides the same API to all databases and since there is no SQL conversion, you can exploit the benefits of each database. SqueakDBX delegates to the C library OpenDBX, which delegates to the C database client library. So...the "worst" thing of SqueakDBX is that you need the OpenDBX library and the database client libraries. This is explained here: http://www.squeakdbx.org/Architecture%20and%20desing Features: http://www.squeakdbx.org/Targets%20and%20Features Documentation: http://www.squeakdbx.org/documentation > How stable/scalable is it? Is > SqueakDBX itself is really stable. It is being used in several production apps without problems. The only tricky (depending on the OS and database) is to install OpenDBX + client libraries. But once it is working, is really stable. > there an ORM on top of it? > > Yes, it is called GlorpDBX, which is a modified version of the Glorp port to squeak (which was all hardcoded to the postgresql native driver). That version is a little old and we only make it work for PostgreSQL, Oracle, MySQL and MSSQL. However, ESUG has sponsored again. The project is being renamed to DBXTalk and the team (the students) are working in two important things: 1) new port (with the latest code) from Glorp of VW to pharo. And try to automatize the process for the future. 2) new mapping tool on top of squeakDBX but much simpler than Glorp. Something like ActiveRecord or SandstoneDB. The main idea for now is to use Magritte for describing the mappings. These two things are just starting.....so you will have to wait ;) -- Mariano http://marianopeck.wordpress.com

May 4, 2011

Fwd: [Pharocasts] New comment on SandstoneDb, simple ActiveRecord style persistence....

by laurent laffont

Someone knows ? Laurent ---------- Forwarded message ---------- From: Bernat Romagosa <noreply-comment(a)blogger.com> Date: Wed, May 4, 2011 at 1:37 PM Subject: [Pharocasts] New comment on SandstoneDb, simple ActiveRecord style persistence.... To: laurent.laffont(a)gmail.com Bernat Romagosa <http://www.blogger.com/profile/04492856362620588522> has left a new comment on your post "SandstoneDb, simple ActiveRecord style persistence...<http://www.pharocasts.com/2010/04/sandstonedb-simple-activerecord-style.html>": I was trying to migrate a simple image-based persistence app into Sandstone, so I just recompiled my biz objects to be subclasses of SDActiveRecord, but of course the objects I want to save are already created and "floating" in the image, so when I try to save them I get an error as many instvars haven't been initialized upon creating these objects (as they were already there). Is there a way to automatically do this? Thanks! Bernat. Posted by Bernat Romagosa to Pharocasts <http://www.pharocasts.com/> at May 4, 2011 4:37 AM

May 4, 2011

VHI (Very Huge Image),

by TedvG TedvG

|question | question = 'Can Pharo run with a VHI (Very Huge Image)? on my Intel i5 Windows7 64-bit 2GB machine Pharo reports with the "space left" command: 7,802,896 bytes (internal) 468,402,192 bytes (physical) 468,402,192 bytes (total) What is in this context: internal and physical? Is this the maximum size of the image on any machine (mine is 2 GB) or does it increase? E.g. is it four times larger if I have 16 GB? What is the maximum image space of Pharo as it is now? theoretically only limited to the address range? 64-bit machines, e.g. an Intel i5 CPU, can handle address range of 2 raisedTo: 64 = 18446744073709551616 bytes That is astronomical, we won't need all that. Anyway, for my Intel i5 661 processor the limit is 16 GB. Waiting for processors with larger memory access.. ' Suppose a CPU can handle a memory of 2 Terabytes and we would need such a large image (all the company's logic and data lives in there) If all the pointers in Pharo are already 64 bit wide?, then such an extremely (at least for today's standards) large image would be possible?. Regards Ted '. pharoHumanBeings do: [ : individual | individual thinkAbout: question; answer; receiveThanks]

May 3, 2011

Re: [Pharo-users] Pharocast ContactManager using SandstoneDb

by laurent laffont

On Tue, May 3, 2011 at 10:09 PM, Doug Edmunds <dougedmunds(a)gmail.com> wrote: > Hello, > > I took the ContactManager code from your Pharocast at > http://www.pharocasts.com/2011/02/pharo-gui-with-polymorph.html > and modified it to run with SandstoneDb, using the ideas from > > http://www.pharocasts.com/2010/04/sandstonedb-simple-activerecord-style.html > > Go ahead and use it for a future Pharocast if you want, or figure out > some place to put it. > > Because Contact is an SDActiveRecord, the database is constructed from > the SandstoneDb records: > > Contacts class >>database > |db| > db := Contact findAll asOrderedCollection. > ^ db > > and so it is portable, so long as the Contacts records are put in the > correctly named Pharo subdirectory. > > > Opening the ContactListEditor uses an initialization which is supposed > to avoid the need to close/reopen Pharo after creating the Contacts class. > > Tested on Pharo 1.3 using SandstoneDb-RamonLeon.135.mcz > > Thanks a lot Doug. May be the next step would be to create a ConfigurationOfContactManager :) Laurent > > -- Doug Edmunds >

May 3, 2011

[COTDC] 58 - TextPrinter

by laurent laffont

Today: TextPrinter Comment Of The Day Contest - One Day One Comment Rules: #1: Each day a not commented class is elected. Each day the best comment will be integrated with name of the author(s). #2: If you cannot comment it, deprecate it. Results: http://code.google.com/p/pharo/wiki/CommentOfTheDayContest Laurent

May 3, 2011