Summary:
Fixes T11784. A lot of things are interacting here, but this probably gets slightly better results slightly more often?
Basically:
- When we load content, we try to keep the viewport "stable" on the page, so the page doesn't jump around like crazy.
- If you're near the top or bottom of the page, we try to stick to the top (e.g., reading the summary) or bottom (e.g., writing a comment).
- But, if you followed an anchor to a comment that's close to the bottom of the page, we might stick to the bottom intead of staying with the anchor.
Kind of do a better job by not sticking to the bottom if you have an anchor. This will get things wrong if you follow an anchor, scroll down, start writing a comment, etc. But this whole thing is a pile of guesses anyway.
Test Plan:
- Followed an anchor, saw non-sticky stabilization.
- Loaded the page normally, scrolled to the bottom real fast, saw sticky stabilization.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T11784
Differential Revision: https://secure.phabricator.com/D17911
Summary:
Ref T12616. Fixes T12715. I suspect these are very rarely used. (I think you tried to get rid of them before but I pushed back since we couldn't really offer great alternatives at the time?)
Now that the code is in a better place:
- Click an inline's header (just the colored part) to select it with the keyboard selection cursor.
- Click again to deselect it.
- You can use "n" and "p" to jump to comments, so "click + n" is the same as the old "V" action.
- This also makes it easier to swap between keyboard and mouse workflows, since you can jump into things with the keyboard at any inline.
Also, make "Reply" render more consistently.
Test Plan:
- Did all that stuff, things seemed to work OK.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12715, T12616
Differential Revision: https://secure.phabricator.com/D17908
Summary:
Fixes T8130. Allows selected comments to be shown/hidden (with "q") or marked done/not-done (with "w").
(These key selections are because "qwer" are right next to each other on QWERTY keyboards, and now mean "hide, done, edit, reply".)
Also, allow "N" and "P" to do next/previous inline, including hidden inlines. This makes "q" to hide/show a little more powerful and a little easier to undo.
Test Plan: Used "q", "w", "N" and "P" to navigate and interact with comments.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T8130
Differential Revision: https://secure.phabricator.com/D17906
Summary:
Ref T12634. Fixes T8131. Currently, most edit operations (edit, reply, collapse, mark done) lose the keyboard cursor state.
Instead, bind the state more tighlty to the inline object itself (instead of the rows which happen to be in the document), and then do a bit of recalculation to try to keep it selected across edits.
Test Plan:
- Used "n" to select an inline.
- Clicked "Done" checkbox.
- Pressed "n".
- Went to the next inline (previously: lost position in document).
- Behavior is also better for: edit, reply, collapse/expand.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12634, T8131
Differential Revision: https://secure.phabricator.com/D17905
Summary:
Ref T12634. Fixes T12633. These events allow the keyboard reticle to resize properly.
(I expect to possibly hide/disable the reticle in the future during edits, but at least make the behavior sensible for now.)
Test Plan:
- Used "n" to select a block.
- Clicked a line number in that block to start a new inline comment.
- Saw reticle resize properly.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12634, T12633
Differential Revision: https://secure.phabricator.com/D17904
Summary: Ref T12634. Fixes T10049. Toggling an inline currently leaves the reticle oddly-positioned.
Test Plan:
- Selected a comment with the keyboard.
- Collapsed it.
- Saw reticle behave reasonably.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12634, T10049
Differential Revision: https://secure.phabricator.com/D17903
Summary: Fixes T12632. Ref T12634. Currently, the keyboard focus reticle does not redraw properly after a window resize.
Test Plan:
- Used "n" to select a block.
- Resized the window.
- Saw the reticle also resize properly.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12634, T12632
Differential Revision: https://secure.phabricator.com/D17901
Summary: Ref T8047. Ref T12634. When we sleep, hide the reticle. Restore it when we wake.
Test Plan:
- With Quicksand enabled..
- Used "j" to select a change in a revision.
- Navigated away by clicking a link.
- WOW! Reticle vanished properly!
- Used "back" to return.
- Reticle returned properly.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12634, T8047
Differential Revision: https://secure.phabricator.com/D17900
Summary:
Ref T12634. Using keyboard shortcuts in Differential currently relies on focus behavior in `KeyboardShortcutManager`.
This possibly made sense long ago, but no longer does, and leads to a whole slew of bugs where the reticle doesn't interact properly with anything else.
Move it to Differential so it can be made reasonably aware of edit operations, Quicksand navigation, etc.
This just moves the code; future diffs will actually fix bugs.
Test Plan: Used "n", "j", etc., saw the same behavior as before.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12634
Differential Revision: https://secure.phabricator.com/D17899
Summary: Ref T12616. This makes "edit" and "reply" work again.
Test Plan:
Used "e" and "r" to edit and reply.
Also used them in bogus ways and got useful UI feedback.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12616
Differential Revision: https://secure.phabricator.com/D17895
Summary: Ref T12616. This moves most keyboard shortcuts into DiffChangesetList. It breaks some shortcuts that I plan to restore later, noted in T12616 (toggle file, edit inline, reply to inline), since I think ripping them out now and rebuilding them in a little bit will make things much simpler.
Test Plan:
- Used j, k, n, p, J, K shortcuts to navigate a revision.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12616
Differential Revision: https://secure.phabricator.com/D17859
Summary:
Ref T12616. This moves "reply" to the new stuff and deletes DifferentialInlineEditor, which no longer does anything.
(This breaks some keyboard shortcuts, but I'll rebase D17859 shortly.)
Test Plan: Replied to inlines; things seemed to work properly.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12616
Differential Revision: https://secure.phabricator.com/D17894
Summary:
Ref T12616. This makes creating inlines use the new code.
Creation and editing is now slightly more consistent in how it uses nodes. This will simplify the next change (replies), which I ran into some trouble with in an earlier iteration.
Note that this (and other changes in the series) allow you to create and edit multiple inlines simultaneously. This is mostly a feature, although I expect we'll need to lock it down a little bit. I have some UI ideas to help avoid errors.
Test Plan: Created inlines on a single line; on a range of lines; on the same line; multiple inlines at the same time.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12616
Differential Revision: https://secure.phabricator.com/D17893
Summary:
Ref T12616. This moves the delete actions to the new, more stateful way of doing things.
These are a little tricky because you can click "Delete" on an inline, but you can also click "Delete" from the preview area at the bottom of the page. If you do, the inline you are deleting may or may not be present on the page.
This has a few bugs -- notably, deleting from the preview without interacting with the on-page inline first won't actually delete the on-page inline yet -- but nothing too serious.
Test Plan: Deleted inlines, undid deletion, deleted from preview.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12616
Differential Revision: https://secure.phabricator.com/D17890
Summary:
Ref T12616. This updates clicking the "Done" checkbox for the new stuff.
This one is pretty clean since the "Done" checkbox doesn't do too much weird magic.
Test Plan: Clicked the box a few times.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12616
Differential Revision: https://secure.phabricator.com/D17888
Summary:
Ref T12616. This doesn't pull over everything (some UI feedback didn't make it yet, and you can't cancel + undo cancelling edits yet) but editing comments technically works.
This is a little shaky, but feels less shaky than every other approach I've tried, so I think I'm finally on a reasonable track here.
Test Plan: Edited some inline comments.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12616
Differential Revision: https://secure.phabricator.com/D17887
Summary:
Ref T12616. Fixes T12153. Currently, when you hide inlines, they hide completely and turn into a little bubble on the previous line.
Instead, collapse them to a single line one-by-one. Narrowly, this fixes T12153.
In the future, I plan to make these changes so this feature makes more sense:
- Introduce global "hide everything" states (T8909) so you can completely hide stuff if you want, and this represents more of a halfway state between "nuke it" and "view it".
- Make the actual rendering better, so it says "epriestley: blah blah..." instead of just "..." -- and looks less dumb.
The real goal here is to introduce `DiffInline` and continue moving stuff from the tangled jungle of a million top-level behaviors to sensible smooth statefulness.
Test Plan:
- Hid and revealed inlines in unified and two-up modes.
- These look pretty junk for now:
{F4948659}
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12616, T12153
Differential Revision: https://secure.phabricator.com/D17861
Summary: Ref T12616. This cements the relationship between ChangesetList (parent container) and Changeset (child) and passes translations down so Changeset can use them to translate the text "Loading..."
Test Plan: Viewed loading changes.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12616
Differential Revision: https://secure.phabricator.com/D17846
Summary: Ref T12616. This ends up being a little messy ("one giant function") and maybe I'll clean it up a bit later, but continue consolidating the wild jungle of behaviors into a smaller set of responsible objects.
Test Plan: Clicked all the menu options, saw them work properly. Grepped for removed methods.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12616
Differential Revision: https://secure.phabricator.com/D17845
Summary: Ref T12616. This class is already mostly-reasonable as a representation of an individual changeset, so I plan to just adjust it a little bit.
Test Plan:
- Used `git grep` to search for `ChangesetViewManager`.
- Used `git grep` to search for `changeset-view-manager`.
- Browsed around and interacted with changesets.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12616
Differential Revision: https://secure.phabricator.com/D17842
Summary: Ref T12616. Put these listeners in DiffChangesetList so the wake/sleep properly for Quicksand.
Test Plan:
- Added some logging.
- With quicksand, moved between diffs.
- Saw "load" and "show more" fire exactly once on each page, with the correct changeset list listener.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12616
Differential Revision: https://secure.phabricator.com/D17841
Summary:
Ref T12616. Ref T8047. Ref T11093. We currently have several bugs where diff state sticks across Quicksand pages.
Add a new top-level object to handle this, with `sleep()` and `wake()` methods. In `sleep()`, future changes will remove/deacivate all the reticles/editors/etc.
See T12616 for high-level discussion of plans here.
This general idea is likely to become more formal eventually (e.g. for "sheets" or whatever we call them, in T10469) but I think this is probably a reasonable place to draw a line for now.
Test Plan:
- Added some logging to sleep(), wake() and construct().
- Viewed changes in Differential.
- With Quicksand on, browsed around; saw state change logs fire properly.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T12616, T11093, T8047
Differential Revision: https://secure.phabricator.com/D17840
Summary:
Ref T11114. Recent changes broke the links to jump to inline comments from the previews because they get hooked up with JS.
Restore the linking behavior.
Test Plan: Clicked "View" on an inline comment preview, jumped to that comment.
Reviewers: chad
Reviewed By: chad
Maniphest Tasks: T11114
Differential Revision: https://secure.phabricator.com/D17131
