April 2014 - Pharo-users - lists.pharo.org

Re: [Pharo-users] Error: No factory specified

by Sven Van Caekenberghe

On 09 Apr 2014, at 03:31, Esteban A. Maringolo <emaringolo(a)gmail.com> wrote: > OpenDBX aside (which I don't plan to use), I don't see how this will > get GLORP running :) It is maintained as part of DBXTalk, as far as I understand http://www.smalltalkhub.com/#!/~DBXTalk/Glorp/ and yes that is a bit confusing, but you can't do much with Glorp unless you can talk to a database, so it makes some sense. > Should I rename Pharo's Cache to something else, like "PharoCache" or > "AbstractCache"? Well, to get the code to load, that would be a dirty hack that could work. But IMO we should rename Glorp's Cache to InstancesCache, it is only referred to via CacheManager. Sven

April 9, 2014

Re: [Pharo-users] Error: No factory specified

by Esteban A. Maringolo

Hi Stephan, 2014-04-08 19:55 GMT-03:00 Stephan Eggermont <stephan(a)stack.nl>: > 'No version found for ''1.3.1'' of ConfigurationOfTalkFFI because: > > Yes, you can see that in > https://ci.inria.fr/dbxtalk/job/DBXTalkDriver-BleedingEdge/BACKEND=pgsql,Pl… > > That's fixed in the configuration I attached. I loaded the configuration you attached, and the loaded the latest version. However it doesn't work (see attached screenshot). It is a fresh 3.0 image with only the config you sent me loaded. OpenDBX aside (which I don't plan to use), I don't see how this will get GLORP running :) Should I rename Pharo's Cache to something else, like "PharoCache" or "AbstractCache"? Thanks in advance, Esteban A. Maringolo

April 9, 2014

Re: [Pharo-users] Error: No factory specified

by Sven Van Caekenberghe

On 09 Apr 2014, at 00:51, Stephan Eggermont <stephan(a)stack.nl> wrote: > "When #totalWeight is no longer below #maximumWeight, the least recently used item of the cache is evicted (removed) to make room. â is not part of the Cache behavior, but of LRUCache. Yes, you are correct: how room is made available and how it is decided which entry to remove is subclass behaviour, but the fact that the cache is limited using an abstract concept of weight is not.

April 8, 2014

Re: [Pharo-users] Error: No factory specified

by Stephan Eggermont

'No version found for ''1.3.1'' of ConfigurationOfTalkFFI because: Yes, you can see that in https://ci.inria.fr/dbxtalk/job/DBXTalkDriver-BleedingEdge/BACKEND=pgsql,Pl… Thatâs fixed in the configuration I attached.

April 8, 2014

Re: [Pharo-users] Error: No factory specified

by Stephan Eggermont

>I disagree: third-party code should not use un-prefixed general names for classes as long as we live in one namespace. That would be nice but creates problems. There is existing third party code we want to use. We can be much more precise in naming, and should not use generic names for specific implementations. "When #totalWeight is no longer below #maximumWeight, the least recently used item of the cache is evicted (removed) to make room. â is not part of the Cache behavior, but of LRUCache.

April 8, 2014

Re: [Pharo-users] Error: No factory specified

by Esteban A. Maringolo

2014-04-08 19:31 GMT-03:00 Sven Van Caekenberghe <sven(a)stfx.eu>: > > On 09 Apr 2014, at 00:16, Esteban A. Maringolo <emaringolo(a)gmail.com> wrote: > >> From my standpoint GLORP is abandoned in the Pharo realm. >> There are some people, like me, that still use it, but no one maintains it. >> It worries me that not a single person ran into this issue before, >> during the whole testing of Pharo 3. Well... I am now, but I'd expect >> more people using GLORP. > > No, it is not abandoned. See > http://dbxtalk.smallworks.com.ar > http://forum.world.st/dbxtalk-working-on-pharo-3-0-td4743313.html DBXTalk is the driver layer, which is not GLORP per se (I know you know this). In the mentioned thread Tudor talks about loading DBXTalk [1], not GlorpDBX. Otherwise he would had run into the Cache name collision. GLORP port is outdated by two years, Alan Knight confirmed it, and Mariano told me porting it from VW is not as straightforward as you would expect (name clashing aside), and the reason for GlorpDBX (which only exists in Pharo/Squeak AFAIK) is to decouple VW's GLORP from PostgreSQL (which is tightly bound). > But your remarks about porting and namespaces is correct. > Maintaining something across multiple Smalltalk implementations is very hard. If not impossible ;-) Thanks for your support. Esteban A. Maringolo [1] I tested the load of DBXTalk in a vanilla 3.0 image using: Gofer new smalltalkhubUser: 'DBXTalk' project: 'DBXTalkDriver'; package: 'ConfigurationOfOpenDBXDriver'; load. #ConfigurationOfOpenDBXDriver asClass loadBleedingEdge. And got he following error: 'No version found for ''1.3.1'' of ConfigurationOfTalkFFI because: Version ''1.3.1'' is not defined in ConfigurationOfTalkFFI. Possible versions include: #(#bleedingEdge #development ''1.0-baseline'' ''1.0'' ''1.1-baseline'' ''1.1'' ''1.2-baseline'' ''1.2'' ''1.3-baseline'' ''1.3'' ''1.4'' ''1.5'')'

April 8, 2014

Re: [Pharo-users] Error: No factory specified

by Stephan Eggermont

There are ci jobs at ci.inria.fr/dbxtalk Iâve started fixing one of them https://ci.inria.fr/dbxtalk/job/DBXTalkDriver-BleedingEdge/ I donât have commit rights The easiest way to make sure glorp keeps working is to create pharo3 (and soon 4) builds for a configuration of all the open source you use.

April 8, 2014

Re: [Pharo-users] Error: No factory specified

by Sven Van Caekenberghe

On 09 Apr 2014, at 00:16, Esteban A. Maringolo <emaringolo(a)gmail.com> wrote: > From my standpoint GLORP is abandoned in the Pharo realm. > There are some people, like me, that still use it, but no one maintains it. > It worries me that not a single person ran into this issue before, > during the whole testing of Pharo 3. Well... I am now, but I'd expect > more people using GLORP. No, it is not abandoned. See http://dbxtalk.smallworks.com.ar http://forum.world.st/dbxtalk-working-on-pharo-3-0-td4743313.html Cache was added recently to Pharo 3. But your remarks about porting and namespaces is correct. Maintaining something across multiple Smalltalk implementations is very hard. Sven

April 8, 2014

Re: [Pharo-users] Error: No factory specified

by Esteban A. Maringolo

2014-04-08 19:04 GMT-03:00 Sven Van Caekenberghe <sven(a)stfx.eu>: > > Argh, yeah, another namespace problem. Sorry you had to run into that. Don't sorry. It was just a matter of trying to load it. > I do wonder though: is there no warning when you are (about to) overwrite an existing class ? Apparently not :D > And I would never name any class of mine anything general like Cache or Record without a namespace prefix. The main code trunk is in VisualWorks, which uses namespaces extensively. So migrating to Pharo/Squeak with an added prefix is not an option, or requires a lot of rewrites during the fileout. > There should indeed be a CI job, I thought there was talk about a DBX one, but I can't find it. I haven't seen it either. >From my standpoint GLORP is abandoned in the Pharo realm. There are some people, like me, that still use it, but no one maintains it. It worries me that not a single person ran into this issue before, during the whole testing of Pharo 3. Well... I am now, but I'd expect more people using GLORP. Esteban A. Maringolo

April 8, 2014

Re: [Pharo-users] Error: No factory specified

by Sven Van Caekenberghe

On 08 Apr 2014, at 23:06, Stephan Eggermont <stephan(a)stack.nl> wrote: > We might be better off renaming Cache to WeightedCache (or NeoCache or SystemCache). > The class comment in Cache seems to cover more than what Cache actually does itself. > > Stephan I disagree: third-party code should not use un-prefixed general names for classes as long as we live in one namespace. I don't understand the remark about Cache being more than what a cache should be: unless it is limited in some way, which any cache should be, you could just as well use a Dictionary. The same argument can be made about the factory concept. Sven

April 8, 2014