I think this is a different issue. If you relocate plugins but you still have all code in plugins/<name>/models/<something>.py then they would work no differently than now (they would be executed in alphabetical order with the plugin name). To do what you ask you would have to put plugin code in modules and import them (something you can already do). The problem is that you would have to be very explicit and import plugins. Installing them would not be enough and that I do not like (at least not in the general case).
On Mar 12, 10:36 am, Thadeus Burgess <thade...@thadeusb.com> wrote: > If all plugins are designed to be class-like, then your example of > plugins just need to inherit. > > The only reason I would be in support for logically changing the > location of plugins is the one of dependencies. > > Meaning, if you have to specify to web2py when to load a plugin, and > in what order... it can handle the dependencies and execute the needed > ones first. > > Dependencies are not hard, its quite simple to write some code to do > this (as I do in py2jquery to handle javascript dependencies). As > Massimo said, the difficulty is in the bytecode compiling. > > I am not convinced there will be any overhead (the slowdown is the 99% > database any ways). It just means web2py will need to be "smarter" > about where stuff is. > > -Thadeus > > On Fri, Mar 12, 2010 at 8:52 AM, mdipierro <mdipie...@cs.depaul.edu> wrote: > > The problem with framework level plugins is that if you pack the app > > and unpack somewhere else then it will not work without the "framework > > level" plugin installed separately. I do not think that is something > > to encourage. Moreover different apps may reply on different versions > > of the plugin and/or need different configuration. > > > Anyway there are two things that can be done at framework level: 1) > > put models in web2py/site-packages 2) framework level models can be > > defined in modules and put there also; 3) framework level views and > > static file can be stored in a new app designed at hoc for this > > purpose and other app can use them too. > > > Massimo > > > On Mar 12, 8:36 am, Alex Fanjul <alex.fan...@gmail.com> wrote: > >> Just to clarify it: > > >> Do we able to conservate my app (rewrited/extended) auth module/model, > >> working alongside "superAuth" thadeus plugin, discarding your framework > >> plugin and system Auth default one? > >> Alex > > >> El 12/03/2010 15:31, Alex Fanjul escribi : > > >> > Ok Massimo, > >> > I agree with you in it makes no sense to rewrite a lot of web2py code. > > >> > Apart from that argument in favor, there is another I don't know if it > >> > would be satisfied right now with plugin_name.py convention: > > >> > -Imagine you write a *framework level plugin* to subsitute auths (or > >> > whatever system feature) views/controllers/models. > >> > -Imagine thadeusb write another *application level plugin* to do the > >> > same called "superAuth" > >> > -Imagine I write an application with an *only modules* extended auth > >> > service with some more fields and stuffs. > > >> > Do we able to conservate my app rewrited/extended auth module/model > >> > over "superAuth" thadeus plugin, discarting system default one? > > >> > Just thoughts, > >> > Alex > > >> > El 12/03/2010 14:01, mdipierro escribi : > >> >> The location of plugins is not a backward compatibility issue. From > >> >> that point of view, we could relocate plugin files. > >> >> The reason I do not want to do is that it is an implementation issue > >> >> that requires rewriting a lot of web2py code (particularly for the > >> >> bytecode-compile functionality), that will make web2py slower, not > >> >> faster, and does not seem to add any new feature (except the > >> >> relocation itself). > >> >> The only argument I have heard in favor of relocation is in fact that > >> >> code will look cleaner with a new plugins location. I do not disagree > >> >> but to users of admin things will look exactly the same (because of > >> >> the logical location according to admin is already the one you > >> >> suggest), to users of the shell models would be scattered and it would > >> >> be more difficult to identify order of execution, and you will get a > >> >> little bit of cleanness is user code at the expense of lots of dirt in > >> >> web2py code (lots of if statements to find out what is where). > > >> >> I will not do it. If somebody wants to write a fully working proof of > >> >> concept implementation to demonstrate that 1) it is not slower; 2) it > >> >> can be done without too much extra complexity in web2py source, I may > >> >> take the patch. > > >> >> Massimo > > >> >> On Mar 12, 4:39 am, Alex Fanjul<alex.fan...@gmail.com> wrote: > >> >>> Hi Massimo, > > >> >>> I haven't said that plugins should have to depend on others, but they > >> >>> should be able to access/play with others to make a trully plugins > >> >>> central network, the dependencies are resoluble at highly level with an > >> >>> exposed convention API like: > > >> >>> plugin_most_active_users.requires=['comments-1.x.x', 'auth-2.x.x'] > >> >>> plugins['tag_cloud'].requires =['tags-1.2.x'] > > >> >>> Its only an idea. > > >> >>> The backward compatibility breaks with the heritance folder structure > >> >>> (as I though you said), isn't it? > > >> >>> *App Level: (example: plugin for commets)* > >> >>> web2py/applications/my_app/plugins/my_plugin/modules/module*.py > >> >>> web2py/applications/my_app/plugins/my_plugin/views/views*.py > >> >>> web2py/applications/my_app/plugins/my_plugin/controllers/controllers*.py > > >> >>> web2py/applications/my_app/plugins/my_plugin/static/statics*.jpg > > >> >>> *Framework Level (example: plugin for ckeditor Editor, or last Wes > >> >>> James > >> >>> coda helper)* > >> >>> web2py/plugins/my_plugin/controllers/controllers*.py > >> >>> web2py/plugins/my_plugin/views/views*.py > > >> >>> The way to ordering load is down-to-up I think, like kohana > >> >>> does:http://docs.kohanaphp.com/general/modules,http://v3.kohanaphp.com/gui.... > > >> >>> Also it's very important the hooks > >> >>> <http://docs.kohanaphp.com/general/events> & events > >> >>> <http://docs.kohanaphp.com/general/events> system, as both of you > >> >>> (thadeus, Massimo) talked at the end of the chat: > > >> >>> There is no calling for new Cache system at all...just was an > >> >>> example... > > >> >>> regards, > >> >>> Alex > > >> >>> El 12/03/2010 5:26, mdipierro escribi : > > >> >>>> I agree with most of what you say. > >> >>>> 99.99% of apps use a single database and so will plugins. This is > >> >>>> because they needs auth to do anything meaningful. > >> >>>> I do not think it is a good idea to have plugins that depend on each > >> >>>> other. dependencies are a mess to manage. In any language and any OS I > >> >>>> ever used. plugins with dependencies are cause for trouble. > >> >>>> But I agree that we may build groups of plugins that cooperate for > >> >>>> some specific tasks (like share access to certain tables and or > >> >>>> certain web services). This will happen for plugins geared toward > >> >>>> specific types of apps so we should not over-engineer it now. > >> >>>> I do not think we need a 2.0 for those things that you asked. We will > >> >>>> get there in small steps and, at this point, I do not see why any of > >> >>>> those improvements should be inconsistent with backward compatibility. > >> >>>> What's your wish list for cache? I never heard anybody calling for a > >> >>>> new cache system. > >> >>>> Massimo > >> >>>> On Mar 11, 9:02 pm, Alex Fanjul<alex.fan...@gmail.com> wrote: > >> >>>>> Very interesting and constructive IRC meeting, congrats to all. After > >> >>>>> reading all text I have some comments: > >> >>>>> - Most of the meeting (50% at least) was concerning about *how > >> >>>>> many and > >> >>>>> what databases should plugins have access to*...it seems the most > >> >>>>> headache for all, BUT, I'm pretty sure that 99% of today real WEB > >> >>>>> applications (and very complex ones) in world uses no more than 1 > >> >>>>> database: think of Magento's, Elgg's, Zimbra's, Active Collab's, > >> >>>>> Twitter's, OpenBravo's, Wordpress's, Drupal's, etc. All of them > >> >>>>> use only > >> >>>>> ONE database (maybe clustered, spreaded, mirrored, etc. but ONE), and > >> >>>>> many of them has very complex plugins systems. The "problem" here, is > >> >>>>> that with web2py its very simple and easy to create a new > >> >>>>> database: just > >> >>>>> do "db=DAL(...)"... and many times we are even "confusing" (in the > >> >>>>> right > >> >>>>> sense) databases with tables... A game for us: Tell me more than 2 > >> >>>>> real > >> >>>>> web applications using more than one database. A reflection: I > >> >>>>> would be > >> >>>>> very afraid if after installing 20 plugins (as I have in my latest > >> >>>>> drupal installation) I bump into 20 (or 15 or 10 or even 5) new > >> >>>>> databases in my phpmyadmin/pgadmin. Yea: be generic and assume all > >> >>>>> posible cases... but.... > >> >>>>> I think Thadeusb was in the right direccion in some > >> >>>>> comments...asumming > >> >>>>> a worst case of ONE shared db for plugins. moreover this would > >> >>>>> simplified things, right? > >> >>>>> - "Turicas: should a plugin access other plugins' data?" --> > >> >>>>> "thadeusb: > >> >>>>> Turicas: I would think no, because a plugin should be self > >> >>>>> contained." > >> >>>>> In this case I disagree, the plugins -for sure- should be able to > >> >>>>> access > >> >>>>> to other plugins data/functions, because as centralplugins grow > >> >>>>> up, many > >> >>>>> of them will be based on others to not reinvent the wheel, so *we > >> >>>>> will > >> >>>>> need a strong convention in exposing API for functions, objects, > >> >>>>> etc.* > >> >>>>> (think of a "plugin_most_active_users" based on thadeus > >> >>>>> "plugin_commets"). > >> >>>>> - Finally I believe that a "heritance folder convention" where you > >> >>>>> can > >> >>>>> override/extend parents functionality/skins/models like the great > >> >>>>> kohana's plugin system (someone mentioned) is the best way to > >> >>>>> achive a > >> >>>>> "easy" and "comprensible" plugin system. Yes, that would suppose a > >> >>>>> big > >> >>>>> change and probably a backward compatible inflexion point, but as > >> >>>>> Massimo said, talk me about functionallity not about implementation. > >> >>>>> Concerning this, and to be honest I'm always thinking of a Massimo > >> >>>>> annunce saying: "Web2py 2.0 Released: the new easier, faster and even > >> >>>>> more powerful python web framework with new DAL, new Plugin > >> >>>>> System, new > >> >>>>> Cache System, new CSS/Form system, etc. (ops but without 1.x backwark > >> >>>>> compatibility sorry)", but it's just a dream :-P > >> >>>>> Is there any new IRC appointment planned? > > ... > > read more » -- You received this message because you are subscribed to the Google Groups "web2py-users" group. To post to this group, send email to web...@googlegroups.com. To unsubscribe from this group, send email to web2py+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/web2py?hl=en.
