Embedding this in index.html won't work anymore, because the Django FastCGI and
the Tornado servers might have been started at different times.
(imported from commit 187909d0593449cf2989857671f9ca526723e451)
Previously if you tried to send to "a b", we actually ended up trying
to send to "a%20b", since we were url-encoding the stream name and
then not properly decoding it.
(imported from commit 307d2999bd309e47fc654ae4422ab4372edde064)
PgUp, PgDn will already call the scroll handler, which will
of its own accord call keep_pointer_in_view.
(imported from commit 8b6e53eaeda91d0f35775de72e16d63dc6c8340b)
That andSelf is only necessary if a .recipient_row were immediately
followed by a .bookend_tr, which, I don't know, after a redesign could
very well be the case, so let's guard against it now, especially since
our performance on scroll is currently pretty good.
(imported from commit 1011555fbfd30943b4aa917997d6e35bcce959fc)
In my limited trial, this sped the call up, on average, from
0.507ms to 0.473 ms... so, admittedly, not a lot.
I think this is a little conceptually cleaner, though, and it handles
the common path with the least work, which I like.
(imported from commit c8b827a2e8111fbdd54bcabe05ac36f64523c466)
nextAll/prevAll walks the entire DOM, basically.
This code only walks the DOM until we find a new .message_row.
This speeds up the average time of a call to this function from about
6.38ms to 0.678ms, in my benchmarking.
Admittedly, the whole outer loop here could still use some
optimization, if we want to; do we really need to call this 1000
times?
(imported from commit 852e2f660a16f8cfd7be35d3271aedb1ac481663)
http://api.jquery.com/first-selector/ mentions that using ":first" in
query selectors can be a little slow since we can't use the browser's
native querySelectorForAll implementation.
In my tests, this cut the average time down from 6.70ms to 6.38ms.
So, not great, but this function is most frequently called
many times in a big while loop, so, better than nothing.
(imported from commit d7725edd0c81431829fe353c6dd2bc61b1de6b19)
If the client is not composing a message, we can just force a page
reload. However, if he is composing a message, we must preserve that
message while still reloading as soon as possible.
We take the following approach: if the client has not completed the
composition after 5 minutes, do a compose-preserving reload
(described below). If he sends the message before the timeout
expires, reload the page after a successful send. If the send fails
(not due to server timeout), however, we do a compose-perserving
reload in case the error was due to the data format changing. If the
send failed due to server timeout, we don't reload because the reload
will probably also fail.
In a compose-preserving reload, we redirect to an URI that has a
fragment indicating we are doing a reload and containing all the
necessary information for restoring the compose window to its
previous state. On page load, we check the fragment to see if we
just did a compose-preserving reload, and, if we did, we restore the
compose window (or just try the send again in the case of send
failure). The URI fragment looks like:
(imported from commit af4eeb3930c24118e088057d4da456748fbd2229)
Given that we now look at the top of the pointer rather than the
bottom for this check, our previous 4/5 number was far too permissive
(and led to your selected message often getting covered up by the
compose window).
Anecdotally, this feels a bit better to me.
(imported from commit 39ca7924db56272f7e623708ce0125b2a7012656)
In particular, we had this issue on huge messages. Suppose you had a
huge message and were scrolling up from the bottom. Initially it would
get selected as the message underneath it was below the
threshold. Then, when you scrolled again we said "oh, well, the top is
too tall, it's also *above* the threshold" and moved the pointer back
down. So it glitched back and forth.
Now we consider you above the view threshold by looking at your
bottom, and below the view threshold by looking at your top, which
seems to resolve this issue.
(imported from commit 2434d1103ab3fa894ade612ac1fb10b2f3cd66cf)
I think this'll allow for a slighlty more accurate drawing of our
floating recipient row in the even that some of these things have
borders and others don't.
(imported from commit 31714f7356604e1d9c64bcc7f6fd14b8a02a99b5)
I'm actually not sure if this is a performance gain or not; I guess it
depends if any of the stuff inside mousewheel or scroll took longer
than 50ms to call (and right now it does).
(imported from commit e3fcc4a14cd8787fa2357a43ed878ab08646e4f2)
This allows us to put other stuff in the narrowbox and have
it also magically get resized. At least in theory.
(imported from commit 92975e3d0893b34d52cad910462cbf2ccaed2eab)
There were two issues:
* The people_list population changed and I failed to noticed
* Typeahead source updating never worked before because calling .typeahead()
more than once does not change the data source
(imported from commit fda14029f4cd37260d82e7bb5689f5022e1b0d28)
This also makes the people_list a list of objects containing the person's full name and email.
(imported from commit cff9b3de8cab0c9b2690ffa60d65d666302b989f)
When Bootstrap shrinks the divs down into 'tablet' mode, our selected
message might end up being even lower in the screen than it was before
(because more text wraps). But our scrollbar does not automatically
advance to keep it in view -- so we do it manually.
This implementation is a little hackish in that it does some
unnecessary recenter_view-ing in big views (which can be a touch
disorienting) to optimize recentering on resizing for small views.
If that behavior is annoying, we can deal with it by being
smarter about when we trigger the recenter code.
(imported from commit 6834e11f7a37833982c388f15174df661d7f55b3)
Without this, if you suspend and resume you wait up to 90 seconds for
long polling to retry.
(imported from commit 5e964c357f395d30107af5b2c934949058c0d3a8)
We had this fascinating behavior where pressing a down arrow near the
end of the page would advance the pointer, call recenter_view, which
would trigger a scroll event, which would call keep_pointer_in_view,
which would notice that we were at the end of the page and advance the
pointer again!
I split out that last part into its own function which is only called
on mousewheel events.
(imported from commit bc85443e762356e3055f8f88585940a1f11f9124)
This also helps us manage checking, case-insensitive, for
subscriptions while preserving the casing used by the class creator
for display.
It also fixes a bug where the class_list would become out of sync with
your true subscriptions, allowing you to appear to send messages to
classes to which you had unsubscribed.
(imported from commit 5e8d017bcfb27a71c52f7517733eda7b926d721b)
So what's now happening here is that the outer div (which we were
previously animating) doesn't get animated at all; instead it is
simply made visible or hidden.
I *believe* this means that it always continues to take the same
amount of space in the DOM; its being added and removed seemed to
cause a slight height-of-page change which caused the scrollHandler to
fire, causing the selected message indicator to move down the page.
So now that this height is fixed, we achieve the animation instead by
animating the inner div -- the one that actually contains the
composebox elements.
(imported from commit 95b2f95d52254e3792633460f42efaac27559efe)
The issue with the animation is that it removes the composebox div
when it's done -- or more relevantly, it "adds" it when the composebox
appears, which causes some DOM elements to get reshuffled slightly
which causes some jitter.
(Similar to what was happening with the email addresses earlier.)
So instead of using display:none, we play with visibility:hidden,
which causes the thing not to show up, but doesn't cause it to
lose its place in the DOM.
(imported from commit a18dbdcd1784b2b54436d48d8425d5fdc8dfbba4)
Rather than trying to keep track of whether the last thing that
happened was an input area being focused (which had all kinds of
bugs), just detect whether we're in an input area using the
appropriate jquery selector. Hopefully this has OK performance.
(imported from commit 6150692ffcb0ab9b04244c3d053b5527847ded2d)
This is a regression introduced by switching from links to buttons,
which get blurred by hide_compose.
(imported from commit f5aa67cb949f3edc7f86a0ef0d7f48b43860138b)
It is confusing, and clicking on it should behave like clicking
elsewhere in the message.
(imported from commit e56434e8e143f6fa58b095e1c7d311b4aa24313f)
The new version is now the only codepath that we use in order to start
a reply to a message.
(imported from commit dd28316d2640fd5fd712f326690d480b7db59c4c)
Just for the record, here's what happened:
If you click on a class message to start a reply, then click outside
to unfocus, then click on a huddle message to reply to that instead,
the system would end up calling the focus handler twice and _then_ the
unfocus handler once (why are we unfocusing something? Because we
have two compose windows -- new_zephyr and new_personal_zephyr, and
we're switching between them). the end result is that we'd re-enable
hotkeys with that unfocus handler.
Fix this by being sure to explicitly disable hotkeys after we setup a
reply to a message.
(imported from commit a7735d9a63f0c3c9f6c12d94e8bb107bf3675f44)
This essentially reverts commit 0d6c3e4d0a70e06b9b2f6e5830b3a9650e44441a,
though also adds support for this behavior with the 'j' key.
(imported from commit 36c51e0f6bd03c53b84abddc75097cb06fc04e16)
Previously we triggered all hotkeys on keydown. Now, if the keydown event
is in the range of a character we bail and wait for it to be handled by the
keypress function instead.
We also redefine all of the keycodes for characters to be their lowercase
versions.
(imported from commit 534199de92174c2858220abb21ce774e1ee0e1d3)
This is somewhat experimental and we may need to work on the condition
when it shows up (or move it elsewhere).
Also, maybe it should say "Today/Yesterday" for times super close to
now -- the main issue with doing this is whether it needs to update
without your reloading the page to avoid being super confusing.
(imported from commit e29faf30c83b9574e5d233213f42a24175f9a616)
This actually is a bit cleaner than our last approach,
because (I believe) we run our scroll handler code
even after the final scroll event, which prevents the
pointer from getting hidden by rapid scrolls.
(imported from commit 3bde4e8f067cd2406f90c04425c6e4ffb81ea784)
This keeps the pointer in the middle 3/5ths of the screen and
asymmetrically handles going above or below those bounds.
(imported from commit 9a6033f53365e6d6cd8f82eb096af849e3b22542)
It causes problems when [un]narrowing, and we already do the rest of formatting
on the server side.
This reverts commit 90af0192b37bbebbf56d5e7c50f182485ddbca10.
This reverts commit df7e355648d2c4d6319de049933547ed96402fd8.
(imported from commit 99d87f0826ec2f49741f86fad6524ed93e76723f)
When you're at the top of the page, you don't get scroll events
anymore if you're scrolling with a mousewheel. So we need
to listen to mousewheel as well as scroll.
(Why not *just* mousewheel? Because then the scrolling done by
PageUp and PageDown no longer causes the pointer to refocus,
because those trigger scroll events.)
(imported from commit 4ee23004f6e090abaabb836f0a9d7b59d0394ced)
There still seems to be some glitching when arrowing around,
which still needs some investigation.
(imported from commit 9c4d8f2b0b55fda5077e2fba70cbe4bcd50b823a)
This code is the start of something that would make your
narrowed-on message remain in the same place post-narrow,
which is possibly a good idea.
(imported from commit 0a557d6492de470928245b33f95ae8c0102db0a6)
I believe we'll get this for free with the new version of
select_zephyr and our bottom_whitespace div.
(imported from commit 0d6c3e4d0a70e06b9b2f6e5830b3a9650e44441a)
It turns out it's marginally easier for us to completely handle Home
and End ourselves rather than let the browser do it, because home
scrolls to the top of the page, but it does not select the topmost
message.
With PgUp and PgDn, the default browser behavior is fine,
but it seems like PgUp on a screen where you're already
"scrolled to the top" should not be a no-op.
(imported from commit 609d639f3a4313e75e1fd6138966f15447b70f2e)
We haven't needed these lately. If we see more problems that need diagnosis
"in the wild" then we can revert this commit.
(imported from commit 3885183d08aefa1f7fd7aa8c2c060e917bf215fb)
I would like to have split this into several commits, but it's hard to do
incrementally, because functions need to be defined before their use in global
data.
(imported from commit 31533396b236f3b657950b073982c317256f082a)
Previously, when we narrowed on a super-long class/instance or
a huddle with lots of people, the box overflowed. This is an
attempt to fix that, even if it isn't the prettiest.
(imported from commit 4eb58726a4c4714bd5435a791ad8fea0eabb58ed)
Previously, connection errors would stick around after a problem with
the server until (1) the server was fixed and (2) we later received a
message. This code change eliminates condition (2) by having the
server return immediately with no messages in the event that the
client has accumulated a large number of errors.
(imported from commit 6a9b08e534db6daea0041a71556ef5b708e935ed)
This fixes a bug where if you unsubbed from a class and then tried to
send a message to it, you could without error and wouldn't see the
message.
(imported from commit 7a4da2a6cb5bbecfc236628dda7b61972b3082e1)
But do keep the pointer on the screen as you scroll, and smoothly
scroll all the way up to the top or all the way down to the bottom.
(imported from commit 6acad2a16339ee5eeeebd62a853000c21e4b7152)
We will probably re-style this eventually.
Also, the animation freezes during template rendering. And the HTML is a "give
up and use tables" situation.
(imported from commit 847374b616dc7ce909834f23d5ed9522aa457254)
Now that it's only called from get_updates_longpoll, we don't need to break out
specific parts that are only needed there.
(imported from commit 60d1cb25d0044ed1b2325d977b869617587e6a0f)
Split off some content previously placed into base back into index.html.
Move content previously localised in accounts_home to a new content_base
for use in the non-app version of the site.
We wrote up some copy and styled it according to taste.
In our new signup form, we no longer collect a short name and instead
extrapolate from the email.
(imported from commit cd734b13a4bda7f304129379524c4cc96b77b24b)
Use 'narrowed' for the same role. (Functions are always true-ish in
Javascript.)
Also, don't add messages to the narrowed table if not narrowed (my bad).
(imported from commit 31707d39abb97472a36d3816c62482e39e732fdc)
Benchmarks [1] show it's much faster than ICanHaz, though we will have to work
to realize those gains.
[1] http://jsperf.com/jquery-template-table-performance/77
(imported from commit b037a6f15e27fd49af88f814ff949811dbcd48f2)
Keegan McAllister
Tim Abbott
Zev Benjamin
Zev Benjamin
Waseem Daher
Jessica McKellar
Luke Faraone