More Changelog

Revision as of 18:06, 29 June 2011 by Sean (talk | contribs) (cat)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

This is the changelog for GSoC project. The project plan and specification is located here.

April, 20 – May, 23 - Bonding period

During this period I got to know better Cliff, Sean and other BRL-CAD comunity members and GSoC students.

I used Drupal forums to gather some information about what others with similar projects have done, what modules have they used and how. It seems that most of similar projects used CCK and views to power most of the site features, with some modules specific to their content (audio, content, etc). Therefore, I plan to use CCK and views, too and to suplement those with the custom modules I'll create for handleing BRL-CAD models task.

During discussions with Sean, Cliff and Erik, some of the features were left out, while others were promoted. Among the most import issues pointed out were (sorted by importance, from most important to less):

  • model licensing
  • import cad geometry (at least .g files, maybe more)
  • export .g files
  • browsing & searching
  • queing operations (conversions, raytracing, etc)
  • multiple upload methods (some model files could be bigger than '1GB')

This period was also good for reading some more information about the many BRL-CAD tools, since I'll have to use many of them in the second part of the project.

May, 25 – July, 10 - Coding period

During this period I have to do initial setup of with Drupal and whatever existing modules I could reuse and select/adapt the site theme of more.brlcad.ord.

May, 23 - June, 2

I got an account on with mysql, SVN, SSH access and all the tools/access rights fast and easy. I started out by installing Drupal and some models: CCK, Views, captcha, flag, pathauto, token.

I decided to use Drupal 5 for various reasons: it's used by, I'm more familiar with it (and contrib modules for it), I'm not very fond of D6, there are more modules available for it. Not all modules I'm using were commited/enabled from the begining; first I like to make sure they work as advertised and that they are useful to me; for others, I didn't know they existed and I added them later.

Also, after a long period of testing and (subjective) comparing many themes, I have finally made up my choice about the site's theme. It's FireflyStream! I like it very much because it's very light (color and code-wise) - my oppinion. I'll probably need to change the logo (maybe one of Cliff's tires).

I also updated codebase for and its modules. During this process, I cleaned up the Drupal modules and themes directories (by moving non-core entries in other places) and installed new modules in sites/all/modules (as recommended by Drupal). This makes later upgrades easier.

With some additional configuration the new captcha module (version 3) is able to add captcha points to the site contact form, something that Sean and I talked about some time ago :)

Meanwhile, I got a database dump and the zipped code from the main site and created a testing environment on my PC. This way I can test things safer and easier and independently of whether my Internet connections is working or not.

June, 3 - June, 10

I tried two ways of implementing licensing: node type and CCK node reference or taxonomy and terms. The advantage of taxonomy solution is its lightweight and the easily ability to produce tag coulds from it. However, it has limited features and posibilities for customization. I'm not sure right now how far this licensing issue will go, so I'd like to have some freedom. As thing will move on, maybe I'll come back to this solution, but for now...

I decided to define a new content type for licenses (in addition to the one I used for models). Then I setup models to require a single node-referenced license. This will force users to select one of the licences we support when submitting a model.

Cliff adviced me to start with some already existing open-source licenses, like GPL. I think it's a good way to do it, because it's easier to add new licences in the future that it could be to drop one license and the reassign its models.

For reporting license problems, first I wanted to use the abuse module. However, it's still in beta and someone advised me to look at a more feature-rich module, called flag.

Using the flag module I defined a new type of tagging, which will allows users to report whenever they notice a license infringement. The administrators can retrive a list of the marked models in order to check the report.

TODO: I think I could integrate this tagging with CIA notifications, so that messages are sent when content is tagged.

June, 11 - June, 16

I installed and configured the pathauto, token, imagefield, filefield and nodequeue modules. These will allow the site to use user friendly addreses. I decided to use the following patterns for content:

  • pages: content/title-raw
  • licenses: license/title-raw
  • models: model/title-raw
  • stories: news/yyyy/mm/dd/title-raw
  • users: user/username

I'll probably need soon the token module for the custom modules I'm working on (to create the model preview images).

The last days I've been checking out some WYSIWYG editors modules for Drupal. There are many out there, although it seems that lately, many of them have converged towards the WYSIWYG API module. Mainly, I focused on testing and comparing browser compatibility and features for the FCKeditor, TinyMCE, HTMLArea (using Xinha), widgeditor, and some others.

The most complete/complex editor was probably TinyMCE. It has many features and plugins. However, once enabled, it seemed to be too much for the simple model descriptions: most users will probably to very little formatting. Also, all these features have with some weights (bigger files and longer loading times). The same thing goes for FCKEditor (which also presented some formatting issues) and Xinha (which has some issues with Opera). On the hand, some other modules (like BUEditor) are too simple and don't seem to help very much in editing your content.

In the end, I decided that widgeditor offers the best balance between features and lightweightness, so we'll stick with it for now.

June, 17 - July, 6

  • June 17 - I tryed to find a module that allows FTP uploaded files to be included in filefields. I found filefield sources module that does something similar, but not exactly, it's still beta and only available for Drupal 6. I also found a forum post from the filefield author saying this feature is not suported and will never be (unless commissioned as a payed job). So I decided to implement it myself by modifing the filefield module. Meanwhile, I also started to commit the other modules I'll use and tomorow I'll copy the database from my PC to the server.
  • June 18 - I finished implementing the local file selection features. This means that model files can be uploaded via the web form or FTP (and later selected). The file selection uses AJAX (when available) to present a list of potential selection. I also finished uploading all the modules. I lost some time struggleling nodequeue because some files had different line endings (Unix, DOS and MacOS) and were rejected.
  • June 19 - I finished bringing everything on the live server. For the theme, I fixed some small bugs. I have created a site logo using the humvee from the BRL-CAD gallery and customized the maintenance page.
  • June 20 - I started working on the custom modules for integration with BRL-CAD.
  • June 21 - June 28 - I was away to Iasi to attend and give a talk at the summer school ECODAM 2009 (focused on applications of Evolutionary Computation in optimisation and data-mining). In the few spare moments I got, I refined the design of the custom modules for integration. They will use mged with custom scripts to perform actions on models.
  • June 29 - I worked on the BRL-CAD code for raytracing models.
  • June 30 - I tested (a bit unsuccessfully) various mged commands to retrieve metadata from models. I seems that mged outputs to data to stderr instead of stdout and when supplying 2 commands separated by ; only the last gets executed. Erik said this looks like bugs. I was unable to get a command to list the trees of the top objects (eval tree [tops] doesn't work). I also worked on an attempt to implement a multistep model submission form. This seems the only approach to allow users to select which objects to render.
  • July 1 - I finally realized that mged actually supports fully Tcl syntax. I started learning basic Tcl and writing the integration scripts for object hierarchy retrieval and metadata retrieval. The custom module will autodetect and support multiple BRL-CAD versions though using these scripts.
  • July 2 - I implemented a connector to PHP through Tcl scripts and redirects, and I started implementing scripts (metadata until now).
  • July 3 - Metadata retriving is completed (tcl scripts and Drupal code). For now I store title, db version, units and summary, but this can be extended with other information since it has the customizable tcl scripts.
  • July 4 - I rewrote the php-mged integration part to use proc_open/proc_close functions instead of exec. This allows more flexibility, including parameter pass between PHP and mged. I implemented the raytracing scripts and code. Last couple of hours I try without success to fix a bug in filefield module that prevents proper cleanup of unused files.
  • July 5 - A lot of things got done today: I fixed the bug in the filefield module; I implemented the queueing process (twice, first with nodequeue, then with flag); I added some tags for models; I optimized and improved the thumbnail generation process (now allows customized and automatically generated images) and added a thickbox gallery for them; I customized some aspects of the theme; I ported everything on and ``Erik helped me by installing the lasted brlcad version.
  • July 6 - I fixed some bugs. I copied everything on and submited two models live for testing. I noticed some more bugs on (raytrace images are created in the wrong directory and title is not retrieved correctly). I fixed the later one.

July 7 - 31

  • July 7 - I started looking at the other features that I have to implement: better browsing/search, better model theming, allow users to select the raytracing objects, multiple formats.
  • July 8 - Today had less time to work, but I focused on the model theming. Especially the teaser of the model should provide enough information to attract the user. This now includes the thumbnail image of the first (which should be more appropriate one, not just the top one).
  • July 10 - Yesterday and today I did some experiments with multistep forms in order to allow users to select which object to show in the images. I'm not very happy with the solution I got so far, so I'll have to work some more on it. The main problem is that multistep submissions are not very easy to implement in drupal. I foresee the same problem (maybe harder) will arrise when implementing the multiple format upload and conversion.
  • July 14 - Last two day I started to implement the object selection for raytracing. In order to do that, I had to reimplement the php-mged interface in order to support multiline outputs. I needed this feature in order to retrieve the list of the objects in the database (and their features). I'm also testing a rewrite of the raytracing feature because after some talks with ``Erik (he mentioned that for realy large databases, loading in mged can take a few minutes). If this works, metadata extraction and raytracing will be done by a single script, starting mged only once per database.
  • July 17 - I finished implementing the new php-mged interface which allows to extract model metadata and create all raytracing of a model with a single call to mged. This should improve performance for raytracing large models or doing many raytraces on a single model.
  • July 18 - I finished implementing object selection for rendering. The user can select which objects from the model file will be drawed before raytracing. In order to do this, I developed a new model that allows displaying multicolumn select lists. I also extended the filefield and imagefield modules to allow the use of the Node Id in the path. This will solve the problem of images being saved in a single directory.
  • July 20-21 - I started working on conversion. According to initial chat with Starseeker and Brlcad, the original file that the user submitted should still be available after conversion. Also, another request is to make the conversion part of the off-line queue-based processing system. I created a new CCK filefield for the original submitted file and now I'm implementing the conversion and system.
  • July 22-23 - I don't know what happened with the activity log for these two days. I'm pretty sure I submitted it and watched it show up, but now I don't see it anymore. Maybe I clicked preview instead of save and then I closed the browser. Basicly I worked on the conversion code.
  • July 24 - Wow! I worked all day, did a lot of code cleanup, refactoring, and in the end, I implemented the conversion from/to BRLCAD file. Works really nice, I'll port it on the server on Monday.
  • July 25 - I added automatic detection for supported file formats. Now, when you create a new file field named forma_file (where format is a supported BRLCAD format), models will be automatically converted to that field. I implemented a new script system and I added fivestar ratings. There is still work to do on the theme and site browsing.
  • July 27-29 - I added various browsing features (by rating, popularity, etc). I implemented a very light download counter for model files, fixed some bugs in conversion code. I had to make some changes to the raytracing code because mged is starting rt in background (which creates problems with PHP IPC functions). I put everything live and started testing.
  • July 30 - I got the cianotify module I developed during the application period and implemented more features and settings. I'll use it to provide feedback in irc about site activity.
  • July 31 - I finished testing the cianotify module. I refined the model processing code so that new raytraced images are produces only if necesary. This reduces the reprocessing times with 80% on models of small-medium complexity.

August 3-Today

  • August 3-6 - I fixed a few bugs. I changed the licence from node to vocabulary because it's lighter and give more features (like tagcloud, models with the same licence), and changed the theming accordingly. I copied everything on production and setup CIA notify.
  • August 7-15 - I posted more example models on the site to check test it some more. I removed the readonly option from mged so that the model titles get displayed correctly (the problem is in the BRL-CAD version installed on the server, but it was fixed in CVS). I added service links and link back code on the model pages so that users can link back to us easily and votes on social sites the models they like. The service links are not displayed for visitors right now (I don't know why, but I'll fix it after GSoC final evaluation, so that I don't break the "pencils down" rule).