What's left to do:
- "context" is now "computation"
- Autorun => Deps.run
- In Deps package, don't say "if we're inside a Deps.run" or anything else that assumes that's the only way to create a computation. Meteor creates computations for you for templates, etc.
This definitely won't "work" because both have API changes.
UglifyJS 1 was having issues minifying certain code (eg lodash). But UglifyJS 2
was occasionally segfaulting (as was other code using large regexps). See the
bottom of https://github.com/laverdet/node-fibers/issues/89
Hopefully this will fix it.
Instead of a general client-side sub de-duping mechanism (which mostly existed
for the sake of autosubscribe, and causes issues with server-driven
unsubscribes), make Meteor.subscribe explicitly aware of reactivity.
Expose an "invalidated" flag on Meteor.deps.Context.
Guarantee that invalidation callbacks from different contexts will not be
interleaved at flush time. This has the implication that if you do
context1.onInvalidate(function () {
context2.invalidate();
});
and this is the only way to invalidate context2, then context2's invalidation
callbacks will not be called until after *ALL* of context1's callbacks are
called. This allows us to be sure that the "unsubscribe, unless autorun tried to
re-create an identical sub" logic runs after the autorun function is rerun.
- Data streamed from the server is quiesced on a per-object basis, not a global
basis. We track which documents a method stubs modifies, and create individual
snapshots ("server documents") of those documents rather than whole-Collection
snapshots. Data writes from the server to documents not modified by stubs are
applied immediately to the local cache; other writes are applied to the
"server document" snapshots. Server documents are flushed to the local cache
when all method stubs that wrote to the document have sent their "data write"
message. (We still do "full database" quiescence after a reconnect.)
- Instead of calling method callbacks as soon as the result is received, we wait
until all data that precedes their "data done" message is flushed to the local
cache. This way, method callbacks can see all of their results locally. (This
applies to Collection mutator callbacks as well.) If this delay is
unacceptable, you can also specify the onResultReceived option to
Meteor.apply; this callback is given the method result as soon as it comes in,
and there's no guarantee that the local cache is up to date.
(This is a client-only change: server-side callbacks do not block on the
write fence.)
- Methods invoked with the "wait" option to Meteor.apply now wait until all
preceding methods are fully finished to be *sent*, not just to call their
callbacks. ie, previous calls block the "wait" method in the same way that
"wait" methods block subsequent calls.
- Remove Meteor.userLoaded and {{currentUserLoaded}}.
Meteor.userId() is now set only at the point where Meteor.user() is fully
loaded.
Current user data is published via an unnamed subscription, not via
"meteor.currentUser".
Replace them with Meteor.loggingIn() and {{loggingIn}}, which become true
as soon as the login method is sent (instead of only once it succeeds).
In accounts-ui, move the spinny into the dropdown, because it now shows up
before error messages would.
- Previously, if we received the "result" message from a method but no "data"
message, and then disconnected and reconnected, quiescence would be
permanently blocked. Now, not only do we allow the app to continue working,
but we even guarantee that the method's callback will be called at the
"reconnect quiescence" point.
- Remove reset function from the Store API (the interface between
_LivedataConnection and Collection), and add a boolean "reset" argument to
beginUpdate instead.
Add saveOriginals/retrieveOriginals functions to the Store API (pass-through
to minimongo implementation).
Allow "replace" messages to be passed to the Store API's update function
(in addition to set/unset).
Allow Store API implementations (eg tinytest_client) to not specify all
functions.
- Server-side tinytest results now stream into the result page instead of
appearing all at once at the end.
- Rename fields and methods of Meteor._LivedataConnection as camelCase, and
prepend all internal fields with _.
- Different Meteor._LivedataConnection objects now have separate
_userIdListeners _ContextSets.
- Remove snapshot/restore functionality from Minimongo collections. (Individual
queries still have result snapshots.) The "server documents" in
Meteor._LivedataConnection serve the equivalent purpose.
- Meteor.loginWithToken's callback is now a "call with error on error, call
with no args on success" callback like the other login callbacks.
- The test-only Meteor._LivedataConnection.onQuiesce function is removed.
Every single use of it is now supported by normal method callbacks.
After reading the Reactivity section for the first time,
I was under the slight impression that a reactive context would
be sufficient for triggering changes. E.g. if the name in the
fragment example were set not from Session, but from a global,
it could be surmised that the HTML fragment would still be
automatically updated if the global variable changed.
This small patch tries to make it clearer that reactivity requires
*both* a reactive context and a reactive data source.
For now, the old names still work as well.
This includes:
- Meteor.isServer/isClient
- this.isSimulation in methods
- Context.onInvalidate
- Meteor.status().retryCount/retryTime
Remove old backwards-compatibility "Sky" alias for "Meteor".
Update all examples in the docs to use camelCase.
Delete unused docs/client/projects.html file.
