Summary:
This diff is designed to dramatically speed up new window load time for
all window types and reduce memory consumption of our hot windows.
Before this diff, windows loaded in ~3 seconds. They now boot in a couple
hundred milliseconds without requiring to keep hot windows around for
each and every type of popout window we want to load quickly.
One of the largest bottlenecks was the `require`ing and initializing of
everything in `NylasExports`.
I changed `NylasExports` to be entirely lazily-loaded. Drafts and tasks
now register their constructors with a `StoreRegistry` and the
`TaskRegistry`. This lets us explicitly choose a time to activate these
stores in the window initalization instead of whenever nylas-exports
happens to be required first.
Before, NylasExports was required first when components were first
rendering. This made initial render extremely slow and made the proposed
time picker popout slow.
By moving require into the very initial window boot, we can create a new
scheme of hot windows that are "half loaded". All of the expensive
require-ing and store initialization is done. All we need to do is
activate the packages for just the one window.
This means that the hot window scheme needs to fundamentally change from
have fully pre-loaded windows, to having half-loaded empty hot windows
that can get their window props overridden again.
This led to a major refactor of the WindowManager to support this new
window scheme.
Along the way the API of WindowManager was significantly simplifed.
Instead of a bunch of special-cased windows, there are now consistent
interfaces to get and `ensure` windows are created and displayed. This
DRYed up a lot of repeated logic around showing or creating core windows.
This also allowed the consolidation of the core window configurations into
one place for much easier reasoning about what's getting booted up.
When a hot window goes "live" and gets populated, we simply change the
`windowType`. This now re-triggers the loading of all of the packages for
the window. All of the loading time is now just for the packages that
window requires since core Nylas is there thanks to the hot window
mechanism.
Unfortunately loading all of the packages for the composer was still
unnaceptably slow. The major issue was that all of the composer plugins
were taking a long time to process and initialize. The solution was to
have the main composer load first, then trigger another window load
settings change to change the `windowType` that loads in all of the
plugins.
Another major bottleneck was the `RetinaImg` name lookup on disk. This
requires traversing the entire static folder synchronously on boot. This
is now done once when the main window loads and saved in a cache in the
browser process. Any secondary windows simply ask the backend for this
cache and save the filesystem access time.
The Paper Doc below is the current set of manual tests I'm doing to make
sure no window interactions (there are a lot of them!) regressed.
Test Plan: https://paper.dropbox.com/doc/Window-Refactor-UYsgvjgdXgVlTw8nXTr9h
Reviewers: juan, bengotow
Reviewed By: bengotow
Differential Revision: https://phab.nylas.com/D2916
Summary:
Up until now, we've been requiring that every plugin control in the composer take the draftClientId, retreive the session, listen to it, build state from the draft, etc. This is a huge pain and is hard to explain to newcomers becaus it frankly makes no sense.
In 0.3.45 we made it so that the ComposerView always has a non-null draft and session. (It isn't rendered until they're available). In this diff, I just pass those through to all the plugins and remove all the session retrieval cruft.
Almost none of the buttons have state of their own, which I think is appropriate.
They do render on every keystroke, but they were already running code (to recompute their state) on each keystroke and profiling suggests this has no impact.
Prepare for immutable
In preparation for Immutable models, make the draft store proxy returns a !== draft if any changes have been made. This means you can safely know that a draft has changed if `props.draft !== nextProps.draft`
Test Plan: Run tests
Reviewers: juan, evan
Reviewed By: juan, evan
Differential Revision: https://phab.nylas.com/D2902
Summary:
- Use a sub-select query with much better performance to display the thread list
- Perform analyze on tables after launch
The new query is:
```
SELECT `Thread`.`data` FROM `Thread` WHERE `Thread`.`id` IN (SELECT `id` FROM `ThreadCategory` AS `M26` WHERE `M26`.`value` IN ('9m9ks71k06n5rmx82kgues09p','9s9k25q6j1krjgpkovbcjm7d','13b7ufruoymvih07ki0uahlto','dtmhlzz6phr47zp512knhjgf8','16dvjb84bszfh15kgfrjj37i3','aclwmgncdqjfibp51bvgbeik','17qad7jhbp6tozog3klm5zagt','4x4bkbawiq825u4eu3aus8tll','7axr9f5f1lzpwm2rw2ghkirhq','dsnn660af0pmou2gg3nstga8a','361qr5rva1ieby2r0ec3sn0bm','10fyvba7pjyjgeyr5i65i1zri') AND `M26`.`in_all_mail` = 1 ORDER BY `M26`.`last_message_received_timestamp` DESC LIMIT 200 OFFSET 0) ORDER BY `Thread`.`last_message_received_timestamp` DESC;
`
0|0|0|SEARCH TABLE Thread USING INDEX Thread_id (id=?)
0|0|0|EXECUTE LIST SUBQUERY 1
1|0|0|SCAN TABLE Thread-Category AS M26 USING COVERING INDEX ThreadFancyIndex
1|0|0|EXECUTE LIST SUBQUERY 2
0|0|0|USE TEMP B-TREE FOR (only on 200 result items)
```
Which is twice as performant as:
```
SELECT `Thread`.`data` FROM `Thread` INNER JOIN `ThreadCategory` AS `M26` ON `M26`.`id` = `Thread`.`id` WHERE `M26`.`value` IN ('9m9ks71k06n5rmx82kgues09p','9s9k25q6j1krjgpkovbcjm7d','13b7ufruoymvih07ki0uahlto','dtmhlzz6phr47zp512knhjgf8','16dvjb84bszfh15kgfrjj37i3','aclwmgncdqjfibp51bvgbeik','17qad7jhbp6tozog3klm5zagt','4x4bkbawiq825u4eu3aus8tll','7axr9f5f1lzpwm2rw2ghkirhq','361qr5rva1ieby2r0ec3sn0bm','10fyvba7pjyjgeyr5i65i1zri') AND `M26`.`in_all_mail` = 1 ORDER BY `M26`.`last_message_received_timestamp` DESC LIMIT 200 OFFSET 0;
0|0|1|SCAN TABLE Thread-Category AS M26 USING COVERING INDEX ThreadFancyIndex
0|0|0|EXECUTE LIST SUBQUERY 1
0|1|0|SEARCH TABLE Thread USING INDEX Thread_id (id=?)
```
Test Plan: Broken!
Reviewers: evan, juan
Reviewed By: juan
Differential Revision: https://phab.nylas.com/D2869
Summary:
The TaskQueue does it's own throttling and has it's own processQueue retry timeout, no need for longPollConnected
Remove dead code (OfflineError)
Rename long connection state to status so we don't ask for `state.state`
Remove long poll actions related to online/offline in favor of exposing connection state through NylasSyncStatusStore
Consoliate notifications and account-error-heaer into a single package and organize files into sidebar vs. header.
Update the DeveloperBarStore to query the sync status store for long poll statuses
Test Plan: All existing tests pass
Reviewers: juan, evan
Reviewed By: evan
Differential Revision: https://phab.nylas.com/D2835
Summary:
This diff implements a behavior change described in https://github.com/nylas/N1/issues/1722.
Reply buttons should prefer to focus an existing draft in reply to the same message, if one is pristine, altering it as necessary to switch between reply / reply-all. If no pristine reply is already there, it creates one.
Reply keyboard shortcuts should do the same, but more strictly - the shortcuts should switch between reply / reply-all for an existing draft regardless of whether it's pristine.
This diff also cleans up the DraftStore and moves all the draft creation itself to a new DraftFactory object. This makes it much easier to see what's going on in the DraftStore, and I also refactored away the "newMessageWithContext" method, which was breaking the logic for Reply vs Forward between a bunch of different helper methods and was hard to follow.
Test Plan: They're all wrecked. Will fix after concept is greenlighted
Reviewers: evan, juan
Reviewed By: juan
Differential Revision: https://phab.nylas.com/D2776
Summary:
This is a critical patch that fixes two problems with the task queue:
1. Tasks in Status: Retry are retried the next time processQueue is run,
which could be pretty much immediately. Certain scenarios lead to tasks
running in a hard loop forever.
2. Returning Task.Status.Retry set the retry flags on the task but did not
schedule the queue to be processed again. So if only a single item in the
queue was present, it might never be retried again until the user performed
another action.
Test Plan: Where did the specs for TaskQueue go? There aren't many... need to write more.
Reviewers: evan
Reviewed By: evan
Differential Revision: https://phab.nylas.com/D2762
Summary:
The old approach we were using to track unread counts by category was really complicated because it involved computing changes to counts in javascript and then syncing them back to the database, from each process that was making queries. Rather than try to fix that, this diff moves us to a new approach where the counts are maintained by executing a query before and after threads are modified to unapply / reapply them from the counters. Doing this in the database in the same transactions as the thread modifications themselves ensures the counts are internally consistent with the Threads table.
This SQL approach is also able to compute initial counts way faster - initializing totals and unreads in a 1GB edgehill.db in about 1 second on my machine.
Test Plan: All old tests removed, new tests coming
Reviewers: evan, juan
Reviewed By: juan
Differential Revision: https://phab.nylas.com/D2757
Summary:
Fix specs
Fix responding to mailto, files at launch
It's super important that `window:loaded` is /not/ sent from index.js because `loadSettings.bootstrapScript` is async and nothing is actually loaded yet. This was causing the app to dispatch the mailto:// links into the main window before a DraftStore existed.
I think this was necessary at one point because we had NylasWindows not using a bootstrapScript? Should not be here anymore...
Test Plan: Run a few new tests
Reviewers: juan, evan
Reviewed By: evan
Differential Revision: https://phab.nylas.com/D2737
Summary:
Previously we always created <blockquote class="gmail_quote"> to wrap quoted text. This is not correct.
Gmail uses blockquotes only when it wants visual indentation, and <div>s to wrap other quoted text, like forwarded
messages which are not displayed indented.
This diff updates N1 to match Gmail exactly. Note that for replies, Gmail actually nests a blockquote.gmail_quote
inside a div.gmail_quote.
I also updated signature handling because it turns out the regexp that was removing existing signatures would blow
away any and all divs until it reached a <blockquote> tag.
Test Plan: See updated specs. Manually tested by creating a thread in Google Inbox and then performing fwd and reply in both N1 and Inbox. Results match.
Reviewers: juan, evan
Reviewed By: evan
Differential Revision: https://phab.nylas.com/D2750
Summary:
This diff replaces "finalizeSessionBeforeSending" with a
plugin hook that is bidirectional and allows us to put the draft in
the "ready to send" state every time we save it, and restore it to
the "ready to edit" state every time a draft session is created to
edit it.
This diff also significantly restructures the draft tasks:
1. SyncbackDraftUploadsTask:
- ensures that `uploads` are converted to `files` and that any
existing files on the draft are part of the correct account.
1. SyncbackDraftTask:
- saves the draft, nothing else.
3. SendDraftTask
- sends the draft, nothing else.
- deletes the entire uploads directory for the draft
Test Plan: WIP
Reviewers: juan, evan
Reviewed By: evan
Differential Revision: https://phab.nylas.com/D2753
Summary:
- Was not properly updating the references to snoozed categories when
accounts were added or removed
- Update whenCategoriesReady to make sure we listen until category syncing has concluded (Move inside CategoryStore)
- #1676, #1658
Test Plan: - TODO
Reviewers: evan, drew, bengotow
Reviewed By: bengotow
Differential Revision: https://phab.nylas.com/D2723
Summary:
This diff also adds an account version number to the config so that the AccountStore can tell whether it should reload accounts (depending on whether it was the instance making tthe changes.)
This diff also fixes a tiny issue where un-opened composers threw an exception if you changed accounts.
Test Plan: New tests
Reviewers: evan, drew, juan
Reviewed By: juan
Subscribers: juan
Differential Revision: https://phab.nylas.com/D2726
Summary:
Previously, we have saved drafts back to the user's provider through the sync engine. There are a handful of very serious edge case issues we're working to solve that are creating a bad user experience. (#933, #1175, #1504, #1237)
For now, we're going to change the behavior of N1 to mitagate these issues.
- If you create a draft in N1, we will not sync it to other mail clients while you're working on it.
- If you enable send later, we'll start syncing the draft to the server as before.
- If you created the draft in another client, we'll sync the draft to the server as before.
Fix specs
Test Plan: Run specs
Reviewers: evan, juan
Reviewed By: juan
Differential Revision: https://phab.nylas.com/D2706
Summary:
Fixes an issue with sending where certain conditions could result in a
duplicated message.
Fixes task dependency logic for draft syncback and send. Changes `createdAt`
on tasks to instead be `sequentialId`, assigned when the task is queued, to
track order of enqueueing. Renames `isDependentTask` => `isDependentOnTask`
and adds comments for clarity.
Test Plan:
Specs updated. Might be good to add some later to test this particular
edge case.
Reviewers: juan, evan, bengotow
Reviewed By: evan, bengotow
Differential Revision: https://phab.nylas.com/D2681
- When accounts changed and the saved perspective could reference
accounts that no longer exist and cause all sorts of errors. This is
fixed.
- Add specs
- Fixes some sentry errors
Summary:
This diff adds an "OutboxStore" which reflects the TaskQueue and
adds a progress bar / cancel button to drafts which are currently sending.
- Sending state is different from things like Send later because drafts
which are sending shouldn't be editable. You should have to stop them
from sending before editing. I think we can implement "Send Later"
indicators, etc. with a simple InjectedComponentSet on the draft list
rows, but the OutboxStore is woven into the DraftList query subscription
so every draft has a `uploadTaskId`.
- The TaskQueue now saves periodically (every one second) when there are
"Processing" tasks. This is not really necessary, but makes it super
easy for tasks to expose "progress", because they're essentially
serialized and propagated to all windows every one second with the
current progress value. Kind of questionable, but super convenient.
- I also cleaned up ListTabular and MultiselectList a bit because they
applied the className prop to an inner element and not the top one.
- If a DestroyDraft task is created for a draft without a server id, it
ends with Task.Status.Continue and not Failed.
- The SendDraftTask doesn't delete uploads until the send actually goes
through, in case the app crashes and it forgets the file IDs it created.
Test Plan: Tests coming soon
Reviewers: juan, evan
Reviewed By: evan
Differential Revision: https://phab.nylas.com/D2524
