Commit graph

760 commits

Author SHA1 Message Date
Ed Sanders 0c2b6c353e Detect collapsed range inside focusable node in ve.ce.Surface#onChange
When the document starts with a focusable node, and we try to initialise
the selection to [1,1], the current logic does not detect that as being
on the focusable node and tries to do a native selection.

Bug: 54446
Change-Id: Ib515c435314c35c4e9b3802da117b963b3ef4169
2013-09-24 16:53:55 +01:00
jenkins-bot 2f4c41ae01 Merge "Add opacity-setting class to GC nodes when generating" 2013-09-23 23:47:13 +00:00
Ed Sanders 3a08aa9b62 Add opacity-setting class to GC nodes when generating
So that they appear 'greyed-out' while updating.

Change-Id: Icbc92d293517beb72d64e4011b0cd5197820de9a
2013-09-23 13:57:56 +01:00
Ed Sanders 936ad149cf Ignore all node attributes in clipboard hash
Some browsers normalise attributes (e.g. Firefox makes
URLs absolute) so we have to remove them from the clipboard
hash to make the comparable.

Bug: 54377
Change-Id: Iadc4d886a5345b28370cbfa31eb665661e577eeb
2013-09-23 12:48:53 +01:00
Ed Sanders 25ae630009 Only listen for copy/paste on documentNode and pasteTarget
As these are the only places we want to override the events.

Bug: 54375
Change-Id: I8a4e207f38508cdaf35cd869f267dd9f1059c83d
2013-09-20 14:48:30 +01:00
David Chan 789d0caf09 Fix pawn on pre-annotation and keypress
modules/ve/ce/ve.ce.SurfaceObserver.js
* pollOnceNoEmit method to update SurfaceObserver's benchmark text

modules/ve/ce/ve.ce.Surface.js
* pollOnceNoEmit from onUnlock

Change-Id: Idb14a6aea723c42109b3825478766799d4abef22
2013-09-11 19:47:39 -07:00
jenkins-bot 71df5562e4 Merge "The Great [...] Rewrite of 2013: Clipboard edition" 2013-09-11 18:15:07 +00:00
David Chan 23921df6dc Refactor CE Surface to reduce event feedback
Restructure SurfaceObserver methods so that the method calls are more precise.

modules/ve/ce/ve.ce.SurfaceObserver.js
* Move timing code into startTimerLoop / stopTimerLoop / timerLoop
* Move poll logic into pollOnce

modules/ve/ce/ve.ce.Surface.js
modules/ve/ui/ve.ui.Surface.js
* Change all surfaceObserver calls to use new API

Change-Id: I0085e4a53c5a776733dce6944b867b8d2228ba4b
2013-09-10 11:54:57 -07:00
Ed Sanders 06c86b3881 The Great [...] Rewrite of 2013: Clipboard edition
Firefox:
* Doesn't fire events without a selection, so when all we have is a
  focusable node, create a dummy selection in the paste target so the
  browser triggers an event on command+C.

Clipboard keys:
* Instead of just creating a text hash which may or may not match up
  with the plain text in the clipboard data, just convert the clipboard
  store into an array and use the index as the key. Also prepend a
  random number (clipboardId) so on paste we can verify the copy came
  from the same CE document.
  The text hash is now used for validation if only HTML is written
  to the clipboard.

HTML:
* Instead letting the browser grab CE HTML, always get DM HTML from
  the converter as it is cleaner (no phantoms, linefeed markers, extra
  class names etc.), and will make cross-ve-instance pasting easier
  in the future.

Clipboard overriding:
* Always override the clipboard with the above HTML, either using
  clipboardData.setData if it is available (webkit) or by throwing
  the HTML into the pasteTarget and selecting it during pre-copy.

Bug: 53375
Bug: 49396
Bug: 52096
Change-Id: Id0f39c10818047bb4b8922e97cead153a770757f
2013-09-10 10:38:34 -07:00
jenkins-bot cba2935173 Merge "Add alt attribute to core image nodes" 2013-09-10 01:43:40 +00:00
jenkins-bot 272b27a959 Merge "Use EventSequencer for CE Surface events" 2013-09-10 01:29:52 +00:00
jenkins-bot 6a81acb44a Merge "Always deep copy when writing to and reading from the fake clipboard" 2013-09-08 03:46:03 +00:00
Ed Sanders dd73b873a8 Add alt attribute to core image nodes
Also update and refactor tests.

Change-Id: I96d08faed42118b713af8011c5d6befb760e65c9
2013-09-07 12:56:17 -07:00
David Chan 79a1bdd691 Use EventSequencer for CE Surface events
Change-Id: I662b164b7bff16ec957f3e06004e477be27282b0
2013-09-06 18:13:44 -07:00
jenkins-bot 0c327d4f73 Merge "Replace renderingEnabled toggle with lock counter" 2013-09-06 23:15:49 +00:00
David Chan 495e83cf2b Replace renderingEnabled toggle with lock counter
modules/ve/ce/ve.ce.Surface.js
* remove methods enableRendering/disableRendering/isRenderingEnabled
* add methods incRenderLock/decRenderLock/isRenderingLocked
* Consequential method call changes
* try-finally blocks around incRenderLock/decRenderLock pairs

modules/ve/ce/ve.ce.ContentBranchNode.js
* Consequential method call changes

Change-Id: Icb71bf536c7bad3a6c0c6512a2efb9565b5bd086
2013-09-06 16:02:51 -07:00
Inez Korczyński 301b287515 Apply DOM changes in onProtectedSetup and onProtectedTeardown instead of constructor
It's important in case of replacing root element ($) of a node, so new element
does have CSS class and contenteditable property set.




Change-Id: Icf14e2164e89114e6b7c10672b782d02f4c5daac
2013-09-05 22:05:23 +00:00
jenkins-bot 27dfe83e6f Merge changes I247d1b68,I5a8ca28a
* changes:
  Fix getOffsetFrom(Element|Text)Node for annotated aliens
  Add test for getOffsetFrom(Element|Text)Node
2013-09-05 07:51:16 +00:00
Ed Sanders 3c24c91a46 Fix getOffsetFrom(Element|Text)Node for annotated aliens
* Replace addOuterLength with a recursionDirection as we only
  want to add length when the first recursion was to the left,
  and add nothing if we recursed to the right.
* Use isContentEditable which is the proper calculated boolean.
  contentEditable is often just (string)'inherit' :/
  Mercifully this works in in IE.
* Only check for the IE ce bug if we aren't recursing at all.

Bug: 53766
Change-Id: I247d1b68484cb510335b5f6789269ec254376cfe
2013-09-05 00:35:53 -07:00
jenkins-bot ae62a0cde8 Merge "Internal lists render nothing in the CE" 2013-09-05 04:56:59 +00:00
Ed Sanders 8f72b083d3 Internal lists render nothing in the CE
Previously they were rendering hidden spans.

Remove hiding from internal items as they have no rendered parent.

Change-Id: Iacc12715900123cd5ab0ddefa713e57e9bea0d1b
2013-09-04 16:49:03 -07:00
Roan Kattouw 2c29dab288 Always deep copy when writing to and reading from the fake clipboard
Otherwise, changes to deeply nested structures like template blobs or
references will leak from the data model into the clipboard data.

Change-Id: I85d8d311f676d45856f30b3747e7383c70c95460
2013-09-04 16:03:56 -07:00
jenkins-bot 630edd049c Merge "Kill domAttributeWhitelist in favor or renderHtmlAttributes" 2013-09-04 00:28:14 +00:00
Trevor Parscal 8dfbc5baa5 Make tools generic and add fancy tool groups
Objectives:

* Got rid of mw prefixing in tools, inspectors and dialogs
* Simplify tool classes so they can be generically used as items in bars, lists and menus
* Add support for a catch-all toolbar group
* Simplify tool registration, leaning on tool classes' static name property
* Move default commands to command registry
* Move default triggers to trigger registry
* Get language tool working in standalone

Change-Id: Ic97a636f9a193374728629931b6702bee1b3416a
2013-09-03 11:27:39 -07:00
Timo Tijhof 8cea089f3b ce: Use a better transparent pixel image
See also http://stackoverflow.com/a/13139830/319266:

> Some are unstable and cause CSS glitches. [If] you have an
> <img> and you use the tiniest transparent GIF possible, it
> works fine[. if] you then want your transparent GIF to have a
> background-image, then this is impossible. For some reason,
> some GIFs such as the following prevent CSS backgrounds (in
> some browsers).
>
> == Shortest (but unstable) ==
> data:image/gif;base64,R0lGODlhAQABAAAAACH5BAEKAAEALAAAAAABAAEAAAICTAEAOw==
>
> == Stable (but slightly longer) *use this one* ==
>
> data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7
>
> Also: don't ommit image/gif. This will break in several browsers.

For the record, this is not limited to rare browsers.
It also affects latest Chrome in some cases as confirmed by
Christian (it'd be white instead of transparent in some cases
when uses as a css background-image without border).

Change-Id: If9ff8a0820c217b6c23e3335944907939a37bef7
2013-08-30 17:51:20 -07:00
Ed Sanders 2d720bd6d9 Allow phantoms and focus areas to be overridden
And actually use this functionality on MWBlockImage to
highlight centred images correctly.

FocusableNode
* Actually use $focusable to render the highlight correctly

ProtectedNode
* Allow a $phantomable element to define which element to
  highlight on mouseenter
* As phatoms are built off shields, make sure a shield is
  given to $phantomable

RelocatableNode
* Allow a $relocatable element to define which element the
  relocatable marker is measured against

ui.Context
* Use $focusable (when available) for positioning the popup

Plus some documentation fixes

Change-Id: I370337239af4fc935cd86757b3ce03011bae5ba8
2013-08-30 14:04:17 -07:00
jenkins-bot 4291b2a308 Merge "When pasting, try to use original range" 2013-08-29 18:26:53 +00:00
jenkins-bot 0ae19ffa40 Merge "Set contentEditable using jQuery#prop" 2013-08-29 17:54:56 +00:00
Ed Sanders 98cad0f410 Set contentEditable using jQuery#prop
Also always use strings 'true'/'false' for the value, because
it is a string property (spellcheck, however, is a boolean)
and always use a capital second 'E' in the name.

Change-Id: Ia4e1f9edfd8f236fe81df190e6a68729d7b52243
2013-08-29 09:59:55 -07:00
jenkins-bot 467ddf5ab8 Merge "Followup 41a4b3fb: use .not()" 2013-08-29 01:55:19 +00:00
Ed Sanders 458f8d1a82 Move contentEditable=false to protected node
Because that's where it should be.

Bug: 53507
Change-Id: I9afffea238a774b884d4d2434d3523e4c4e445b2
2013-08-28 16:18:55 -07:00
Ed Sanders e0018e8e94 Rename 'DOM Changes' => 'DOM changes'
Because we don't use title case in comments

Change-Id: I5cdd3241aba0727a3cd5d0ded972ae54abf1efd0
2013-08-28 15:55:35 -07:00
Ed Sanders 182d9bbeea Check clipboardData exists before accessing it
Bug fix for browsers which don't have an event.clipboardData
property. The check is done correctly in onCopy but not onPaste.

Change-Id: I3cdf4c0358aa145dcb70c2c103d08a0002001fd4
2013-08-28 11:47:13 -07:00
Ed Sanders ddb608e6bc When pasting, try to use original range
This functionality was (accidentally?) removed by I8f8a240a. It
stops us from over-zealously balancing data when we can get away
with just pasting the original selected range (e.g. it always wraps
collections of text and content nodes in another paragraph).

Bug: 53364
Change-Id: I93fa56c4e43083993c310e0050087e9d1de1e08b
2013-08-27 15:40:59 -07:00
jenkins-bot 92c7be9b31 Merge "Detect outdated pending post KeyPress handler" 2013-08-27 22:31:07 +00:00
David Chan d867ef8fc0 Detect outdated pending post KeyPress handler
modules/ve/ce/ve.ce.Surface.js
* Schedule the post-keypress async handler in a way that can be cancelled.
* Cancel it and run the handler immediately if another key event happens.

Bug: 53079
Change-Id: If139ff3230c10caa616743f71659c4606d290310
2013-08-27 13:55:02 -07:00
Roan Kattouw 10393b8773 Followup 41a4b3fb: use .not()
I claimed that it didn't preserve text nodes, but it does. It's just
that there's a bug where it doesn't preserve text nodes if the selector
passed to it is too simple, but 'link, meta, style' is complex enough
to avoid triggering this bug.

Change-Id: Id4a60dc87b8d4c613bc7013641b116dd7c331ac1
2013-08-27 13:53:17 -07:00
Roan Kattouw 051867c3ab Actually use GeneratedContentNode in AlienNode
ce.AlienNode was mixing in GeneratedContentNode, but wasn't actually using
it properly. This commit fixes this so that AlienNode can benefit from
the <meta>/<link>/<style> stripping code in GeneratedContentNode.

Also added a getHashObject() function to dm.AlienNode to summarize the
domElements attribute because ve.getHash() chokes on it otherwise.

Change-Id: Ief3be94f9730297abe0e3c57506b81a8ff1d136d
2013-08-27 13:53:17 -07:00
jenkins-bot 8b773cf9a4 Merge "Give document real focus after cut" 2013-08-27 20:21:42 +00:00
jenkins-bot 80ded50c8b Merge "Copy-paste data fixes" 2013-08-27 20:21:00 +00:00
jenkins-bot 78141a6f33 Merge "Don't render <meta>/<link>/<style> tags in GeneratedContentNode" 2013-08-27 18:28:34 +00:00
Roan Kattouw 41a4b3fb51 Don't render <meta>/<link>/<style> tags in GeneratedContentNode
There is no good reason why we'd render these tags, and their presence
causes Chrome to crash while cutting across them.

Bug: 50043
Change-Id: I611e3907cf20fa27dbef89ea941d0b787a44ba4f
2013-08-27 11:26:37 -07:00
Roan Kattouw ce138adbdb Kill domAttributeWhitelist in favor or renderHtmlAttributes
domAttributeWhitelist wasn't actually being observed. Instead, we
already had the code in place to treat renderHtmlAttributes as a
boolean/whitelist/blacklist hybrid, it just wasn't used that way
yet.

This makes the interface nicer and fixes the bug where all attributes
(including data-parsoid and even things like onmouseover) were
rendered by CE.

Change-Id: I02e266c7c7dc197ed845164b7a705d786846a33b
2013-08-26 18:37:50 -07:00
Ed Sanders fd410c03a7 Give document real focus after cut
If only a FocusableNode is selected the document doesn't have
real focus, so we must do it manually.

Bug: 53362
Change-Id: I781f59dda7f2884ff02f6688c2c8c037ac7ed1a6
2013-08-26 23:06:13 +00:00
Ed Sanders 99692230b1 Copy-paste data fixes
* Assign cut/copy/paste events to the document so that
  they are triggered when the CE doesn't have real focus
  e.g. only a FocusableNode is selected
* Use clipboardData.setData when available. The key can
  be stored and retrieved perfectly in text/xcustom but
  this technique doesn't work in FF or IE
* Just use text content for key as this is what is written
  to the clipboard, so this can be used directly without
  having to get the content from the pasteTarget.

Bug: 48604
Bug: 49396
Change-Id: Ib3702f9441f6ee3fa34ec071f00994dd7e591d99
2013-08-26 15:25:15 -07:00
jenkins-bot 97c03da395 Merge "Don't emit Surface changes back to the Surface" 2013-08-22 09:48:23 +00:00
David Chan 24dae99065 Don't emit Surface changes back to the Surface
modules/ve/ce/ve.ce.SurfaceObserver.js
* add emitContentChanges argument to poll (and start and stop)

modules/ve/ce/ve.ce.Surface.js
* Pass emitContentChanges=false to calls to start/stop in certain places
* Explicitly pass emitContentChanges=true elsewhere, to preserve prior
  behaviour

modules/ve/ui/ve.ui.Surface.js
* Explicitly pass emitContentChanges=true to poll

modules/ve/test/ve.test.js
* Escape non-BMP unicode text literals (to prevent potential editor
  problems)

Bug: 50105
Bug: 50346
Bug: 50631
Bug: 51477
Bug: 52716
Change-Id: I19ec7eaa0e3224cbfc7e7188e964183d7393c9a7
2013-08-21 17:26:32 +00:00
jenkins-bot 4f6e42b699 Merge "Subscript is one word (and so is superscript for that matter)" 2013-08-20 20:37:49 +00:00
Trevor Parscal 4e6694c75e Subscript is one word (and so is superscript for that matter)
Objective:

* Unify the capitalization of subscript and superscript

Change-Id: I5a7fbc2d864c896ebb26177cf05c7adddacd8829
2013-08-20 13:34:50 -07:00
David Chan 1987d3aa2b Don't emit Surface changes back to the Surface
modules/ve/ce/ve.ce.SurfaceObserver.js
* add emitChangeEvents argument to poll (and start and stop)

modules/ve/ce/ve.ce.Surface.js
* Pass emitChangeEvents=false to calls to start/stop in certain places
* Pass async=false in the onDocumentKeyDown call to stop
* Explicitly pass emitChangeEvents=true elsewhere, to preserve prior behaviour

modules/ve/ui/ve.ui.Surface.js
* Explicitly pass emitChangeEvents=true to poll

modules/ve/test/ve.test.js
* Escape non-BMP unicode text literal (to prevent potential editor problems)

Change-Id: Ia96efad0c808b9a02e508bbfda65fdf4f5cedeff
2013-08-19 14:20:08 +00:00