While trying to get translation and citing working with asynchronously
generated data, we realized that drag-and-drop support was going to
be...problematic. Firefox only supports synchronous methods for
providing drag data (unlike, it seems, the DataTransferItem interface
supported by Chrome), which means that we'd need to preload all relevant
data on item selection (bounded by export.quickCopy.dragLimit) and keep
the translate/cite methods synchronous (or maintain two separate
versions).
What we're trying instead is doing what I said in #518 we weren't going
to do: loading most object data on startup and leaving many more
functions synchronous. Essentially, this takes the various load*()
methods described in #518, moves them to startup, and makes them operate
on entire libraries rather than individual objects.
The obvious downside here (other than undoing much of the work of the
last many months) is that it increases startup time, potentially quite a
lot for larger libraries. On my laptop, with a 3,000-item library, this
adds about 3 seconds to startup time. I haven't yet tested with larger
libraries. But I'm hoping that we can optimize this further to reduce
that delay. Among other things, this is loading data for all libraries,
when it should be able to load data only for the library being viewed.
But this is also fundamentally just doing some SELECT queries and
storing the results, so it really shouldn't need to be that slow (though
performance may be bounded a bit here by XPCOM overhead).
If we can make this fast enough, it means that third-party plugins
should be able to remain much closer to their current designs. (Some
things, including saving, will still need to be made asynchronous.)
This mostly gets ZFS file syncing and file conflict resolution working
with the API sync process. WebDAV will need to be updated separately.
Known issues:
- File sync progress is temporarily gone
- File uploads can result in an unnecessary 412 loop on the next data
sync
- This causes Firefox to crash on one of my computers during tests,
which would be easier to debug if it produced a crash log.
Also:
- Adds httpd.js for use in tests when FakeXMLHttpRequest can't be used
(e.g., saveURI()).
- Adds some additional test data files for attachment tests
Also:
* _finalizeErase in Zotero.DataObject is now inheritable
* Call _initErase before starting a DB transaction
* removes Zotero.Libraries.add and Zotero.Libraries.remove (doesn't seem like this is used any more)
There's a lot more to do, and this isn't ready for actual usage, but the
basic functionality is mostly in place and has decent test coverage. It
can successfully upgrade a library last used with classic syncing and
pull down changes via the API. Uploading mostly works but is currently
disabled for safety until it has better test coverage.
Downloaded JSON is first saved to a cache table, which is then used to
populate other tables and later for generating PATCH requests and
automatically resolving conflicts (since it shows what was changed
locally and what was changed remotely). Objects with unmet dependencies
or unknown fields are skipped for now but don't block the rest of the
sync.
Some of the bigger remaining to-dos:
- Tests for uploading
- Re-do the preferences to get an API key
- File sync integration
- Full-text syncing integration
- Manual conflict resolution (though this already includes much smarter
conflict handling that automatically resolves many conflicts)
And use it in resetDB() test support function, mainly to allow
skipBundledFiles for resetDB calls. Translator installation and
initialization can take a long time, but tests that need a clean DB
don't necessarily rely on translators. Without this, running resetDB()
in beforeEach() for many tests is prohibitively slow.
Since modal windows (e.g., the Create Bib window and the Quick Copy site
editor window) can't use yield, style retrieval
(Zotero.Styles.getVisible()/getAll()) is now synchronous, depending on a
previous async Zotero.Styles.init(). The translator list is generated in
the prefs window and passed into the Quick Copy site editor, but it's
possible the translators API should be changed to make getTranslators()
synchronous with a prior init() as well.
- Simplified schema
- Tags are now added without reloading entire tag selector
- On my system, adding 400 tags to an item (separately, with the tag
selector updating each time) went from 59 seconds to 42. (Given that
it takes only 13 seconds with the tag selector closed, though,
there's clearly more work to be done.)
- Tag selector now uses HTML flexbox (in identical fashion, for now, but
with the possibility of fancier changes later, and with streamlined
logic thanks to the flexbox 'order' property)
- Various async fixes
- Tests
Get rid of data_access.js, at long last. Existing calls to
Zotero.getCollections() will need to be replaced with
Zotero.Collections.getByLibrary() or .getByParent().
Also removes Zotero.Collection::getCollections(), which is redundant
with Zotero.Collections.getByLibrary(), and Zotero.Collections.add().
The latter didn't didn't include a libraryID anyway, so code might as
well just use 'new Zotero.Collection' instead.
Groups were already being loaded for the collections list, so we might
as well just store them initially and let Zotero.Libraries.getName() be
a synchronous call.
'b' for *b*undled files
Translators and styles take a long time to install and initialize in
source installations, and they're unnecessary for many tests.
This shaves about 10 seconds off each test run for me on one system (and
that's with some help from filesystem caching).
0 allowed for some accidental behavior due to old code expecting NULL,
and it prevented easy checks (``if (!libraryID)``) for a passed
libraryID. Code now uses Zotero.Libraries.userLibraryID instead of a
hard-coded value (except in schema.js). Functions can still make
libraryID optional, but they should then use
Zotero.Libraries.userLibraryID if that's to mean the user library.
There might be some code that still expects 0 that I missed.
- Protocol handler extensions can now handle promises and can also make
data available as it's ready instead of all at once (e.g., reports now
output one entry at a time)
- zotero:// URL syntaxes are now more consistent and closer to the web
API (old URLs should work, but some may currently be broken)
Also:
- Code to generate server API, currently available for testing via
zotero://data URLs but eventually moving to HTTP -- zotero://data URLs match
web API URLs, with a different prefix for the personal library (/library vs.
/users/12345)
- Miscellaneous fixes to data objects
Under the hood:
- Extensions now return an AsyncChannel, which is an nsIChannel implementation
that takes a promise-yielding generator that returns a string,
nsIAsyncInputStream, or file that will be used for the channel's data
- New function Zotero.Utilities.Internal.getAsyncInputStream() takes a
generator that yields either promises or strings and returns an async input
stream filled with the yielded strings
- Zotero.Router parsers URLs and extract parameters
- Zotero.Item.toResponseJSON()
This required doing additional caching at startup (e.g., item types and fields)
so that various methods can remain synchronous.
This lets us switch back to using the current Sqlite.jsm. Previously we were
bundling the Fx24 version, which avoided freezes with locking_mode=EXCLUSIVE
with both sync and async queries.
Known broken things:
- Autocomplete
- Database backup
- UDFs (e.g., REGEXP function used in Zotero.DB.getNextName())
Promise-based rewrite of most of the codebase, with asynchronous database and file access -- see https://github.com/zotero/zotero/issues/518 for details.
WARNING: This includes backwards-incompatible schema changes.
An incomplete list of other changes:
- Schema overhaul
- Replace main tables with new versions with updated schema
- Enable real foreign key support and remove previous triggers
- Don't use NULLs for local libraryID, which broke the UNIQUE index
preventing object key duplication. All code (Zotero and third-party)
using NULL for the local library will need to be updated to use 0
instead (already done for Zotero code)
- Add 'compatibility' DB version that can be incremented manually to break DB
compatibility with previous versions. 'userdata' upgrades will no longer
automatically break compatibility.
- Demote creators and tags from first-class objects to item properties
- New API syncing properties
- 'synced'/'version' properties to data objects
- 'etag' to groups
- 'version' to libraries
- Create Zotero.DataObject that other objects inherit from
- Consolidate data object loading into Zotero.DataObjects
- Change object reloading so that only the loaded and changed parts of objects are reloaded, instead of reloading all data from the database (with some exceptions, including item primary data)
- Items and collections now have .parentItem and .parentKey properties, replacing item.getSource() and item.getSourceKey()
- New function Zotero.serial(fn), to wrap an async function such that all calls are run serially
- New function Zotero.Utilities.Internal.forEachChunkAsync(arr, chunkSize, func)
- Add tag selector loading message
- Various API and name changes, since everything was breaking anyway
Known broken things:
- Syncing (will be completely rewritten for API syncing)
- Translation architecture (needs promise-based rewrite)
- Duplicates view
- DB integrity check (from schema changes)
- Dragging (may be difficult to fix)
Lots of other big and little things are certainly broken, particularly with the UI, which can be affected by async code in all sorts of subtle ways.
Cmd on OS X, Shift on Windows/Linux
How do I not get to close a ticket for this?
Unfortunately on Windows it doesn't seem possible to set the cursor
effect to arbitrary states (see note in libraryTreeView.js::
_setDropEffect() for the gory details), so this just uses the default
cursor there. On OS X and Linux the cursor reflects the requested
action.
This should make it much easier to debug startup errors, particularly in
Standalone.
This also adds a general mechanism to set Zotero initialization options via
command-line flags.
- Close Zotero pane before database is closed prior to reload, instead
of waiting until reload is complete
- Show an error message if Zotero Standalone is not accessible when it
should be
