* Wordpress install process needs to be fixed also: it's easy to forget to define what hostname that the wordpress plugin should use. This should be an option in the wordpress UI, instead of something that is hardcoded * tests should be yamlized too, and need clearer error messages. * mobile width - detail - paradigm disappears * need to hide 'more information' link when there isn't actually more information - probably slightly easier now * some tags are too long for the field, but overflow is hidden - f.eks.) http://localhost:5000/detail/sme/nob/goabbat%20guoibm%C3%A1me.html?no_compounds=true&lemma_match=true -- pron. resipr. 1.p.tot. Acc/G ... -- # General design stuff example from detailed view: - lexicon( xml -> parser ) + language specific formatting (chance for markup) - - user query -> view -> lexicon lookup results -> formatter -> view does some more things, sorting, object manipulation -> templates - templates obviously handle markup this is too much, but it happened because a lot of functionality was developed on the fly. now that all the needs from the lexicon and morphology are clear, it needs to be improved and simplified. - morphology needs: - - lexicon needs: - lexicon-specific formatting must be turned into something the application can display. - lexicon-specific tags must be turned into something morphology can operate on without much effort from a programmer - room for entry/word-class specific formatting stuff - linguist needs: - choose how to display a specific entry type based on knowledge of lexicon and morphology, without much effort from the programmer - pos=N, type=Prop and has @reg; -> display l/text() with l/@reg - pos=N, type=Prop and has @reg; -> display l/text() with l/@reg - able to programmatically define links to other entries - able to define what paradigm to generate and show with the entry - interact with morphology and tags without having to know how all the (python) code works # Misc * update reversed to work with new structures -> fin-kpv * ie8 rendering bug on flag lists in header on main page, but not on about page * detail view, include generation info in opt+shift+? debug * See bugs * change cache settings for plugin js in nginx configs, some browsers are unwilling to get a new version even when it exists. * refresh settings button in plugin * cakefile include build bookmark.min.js * a little advertising and such for the reader and page with content in specific languages see script in misc/o Add text to reader page: "Here are the top websites that people are reading, did you know you can read them too?" 14 www.ntfk.no 14 www.skogholt.org 15 yle.fi 16 www.sametinget.se 17 fi.wikipedia.org 19 dagogtid.no 20 www.facebook.com 23 www.samediggi.no 25 samediggi.fi 25 sametinget.se 28 samisk.karasjok.kommune.no 29 www.nrk.no 30 www.hs.fi 44 sanit.oahpa.no 46 giellatekno.uit.no 46 www.regjeringen.no 56 se.wikipedia.org 63 nrk.no 161 avvir.no 277 skuvla.info 582 www.avvir.no 3429 kursa.oahpa.no For privacy concerns, only include top _domains_ not individual pages. Also, split these out by language pair used in the lookup. above list is just everything in access logs. ## post launch TODOs: * bookmark icon for iOS /apple-touch-icon-57x57-precomposed.png /apple-touch-icon-57x57.png /apple-touch-icon-precomposed.png /apple-touch-icon.png * gtweb: certificates * translate the mobile bookmark bubble, maybe modify/write one for android, where the process is a little more complex * Link til read i selve ordbok-grensesnittet -- kor? -- "Vil du ha oversatt ord i en tekst?" -- "Háliidatgo jorgalahttit sániid teavsttas?" -- "Haluatko käännöksen tekstissä esiintyvään sanaan?" * autocomplete SoMe variant ## Next: * Glossary page explaining each tag part. * mobile menu language variant display (e.g. SoMe vs. Norm) * bookmarklet unload on additional click * word test-list - http://giellatekno.uit.no/doc/dicts/checklist.html * dropdown events on mobile are a bit funny for the menu on the detailed view, and it is impossible to select a new language pair * use logger's own log message formatter -- * Once all this is together, optimize memory usage, with specific attention to gtlab / gtoahpa, may be a better way to have multiple threads/greenlets/etc running while parsing trees into only one thread. Note: http://blog.pythonisito.com/2012/07/gevent-and-greenlets.html ## Would be nice * Makefile: check which FST and lexica are updated on svn up, and then automatically rebuild all * developer- and linguist-friendly templates for editing how words are presented ; multimethods + jinja - prefer more templates instead of many if/else statements for attributes? * tagsets for selecting POS from analyses & specifying what parts of tags should be used for lookups, with fallbacks if they don't exist. (i.e., take POS and Valence if both exist in analysis, if only POS, use that) - sma: +1+V, +2+V; one fix is to remove from FST, so that +V is first element, other option would be to improve the lookup system. - maybe thinking about a solution to this will help generalise other things for more languages, allowing language-specific configuration of what elements from the analysis must be used to look up words in the lexicon * advanced search form: here we get to have some nerdy linguistic fun. ## Features to develop at some point: * Use something other than werkzeug's default cache * Maybe should provide an individual process for each language pair? Should be easy to split off new processes by pair/endpoint. Note also werkzeug DispatcherMiddleware and PathDispatcher: http://flask.pocoo.org/docs/patterns/appdispatch/ * getting near the time to adapt version of lookupserv.py that is checked in for actual use. spawn it on process run, so that no one needs to think/worry that it exists aside from looking at process hierarchy in htop