Table of Contents
VuFind Developers Call Minutes: April 1, 2014
Attending: Filipe Bento, Chris Hallberg, Demian Katz, Benjamin Mosior, Tod Olson, Ben Wiens
1. Development Updates
- VUFIND-262 - This ticket has been closed now that VuFind's Solr spelling processing logic has full test coverage.
- VUFIND-442 - This ticket suggested providing sample code for building a custom compiled SolrMarc indexer, but the new Mixin functionality makes such a solution unnecessary. Demian has added some better Mixin documentation and closed the ticket.
- VUFIND-827 - The cover loader now supports ISSNs, with support implemented for Syndetics.
- VUFIND-848 - Demian implemented better support for title highlighting in the Solr module (to solve the problem of snippets causing the beginnings of titles to be removed).
- VUFIND-966 - This ticket suggests cleaning up controllers to eliminate direct service manager access.
- VUFIND-967 - Chris opened this ticket to track improvements to tablet mode in the Bootstrap theme (previously, tablets went to “phone mode” which looked strange on the larger displays); he has made some changes but is open to further improvements.
- VUFIND-968 - This ticket tracks a problem with grid view and combined view not playing well together. It is not yet resolved.
- VUFIND-969 - This ticket tracks a requested feature to load ALL facet values in the side facet area for selected fields.
- #125 - This PR contains work in progress on expanded holds functionality.
Several additional PRs have been opened and merged recently resolving various small bugs (mostly in the Bootstrap theme). Most serious: #124, which fixes a problem with “cancel selected holds” in the Blueprint theme.
2. Development Planning
Zend Framework 2.3
ZF 2.3 raises the PHP requirements. Demian proposes not upgrading to the new framework until after VuFind 2.3 to give people time to upgrade, but raising the requirement to PHP 5.4 for VuFind 2.4 and up.
Nearly everyone on the call still has PHP 5.3.3 running in production, so we should review this situation as time goes on.
Testing has been bumped from 1 million to 12 million documents (using Harvard data) for a more realistic high-end environment. Tried 1, 2 and 3 GB heap settings – all were overwhelmed within a few hours of jMeter hammering (out of memory caused by filter cache filling). Tests used generic searches which might have caused unusually large filter cache contents – however, this still seems to confirm that current settings are too high. Should be 4 or 5 digits. Autowarming is also a very long process that puts a significant load on the server due to large autowarm counts. Everything needs to be scaled down.
A message to the Solr mailing list also confirmed the theory that filter cache size should be reduced.
Next step: determine better numbers (and possibly test in production).
Call Number Normalization
Not much progress since last time, but a couple of things:
1. Tod talked to Bob Haschart at Code4lib about a variety of SolrMarc improvements; Bob suggested posting a summary to the solrmarc-tech list for discussion. Tod suggests anyone interested in the future of SolrMarc should pay attention to the list in the near future to join in the conversation.
2. Another use case has come up for index-specific normalization: insertion/removal of hyphens for subject subdivisions in headings (which needs to be handled differently for display vs. searching/ordering). Tod will put together a pull request for his topic normalizer.
Ben and Chris have discussed this a bit, but Code4lib prevented major progress. They will continue discussing.
Ben mentioned that the “library OPAC meets Web 2.0” tagline is not aging well; we should try to brainstorm a new one to go with the new logo.
4. Code4lib Wrap-Up
Chris shared some breakout session notes. Tod mentioned some discussion of OLE support – Dave Lacy is working on a driver right now.
5. Next Call Date
Demian will be traveling on the next scheduled call date. We'll go ahead anyway with Chris taking the lead.
6. Other Topics?
Chris has (for April 1) implemented a VuFind Multiplayer Mode which uses a node.js server to allow users to see each other's cursors moving around when accessing the same pages. Feel free to merge in if you wish to confuse people for a day!
The next call will be Tuesday, April 15, 2014 at 10am Eastern Daylight Time (14:00 GMT).