Before this commit, most of the time, components are destroyed when the
virtual dom is patched and a component node is removed. However, since
Owl is asynchronous and a component may take some time to get ready
(with onWillStart), it can happen that a component is created, then
before it is ready, it is recreated. In that case, the initial instance
has to be destroyed.
Before this commit, the destroy operation was done immediately, when we
cancel the current fibers. However, this means that we cannot have a
guarantee between micro task ticks that a component has not been
destroyed in the meantime.
For example, in Odoo, it is common to use the rpc service, which will
throw an error if called by a destroyed component. But because of the
possible destruction of a component at any time, the following code is
unsafe:
async loadSomeData() {
// guaranteed to be called when component is alive
await Promise.resolve();
// however here, component may have been destroyed
this.rpc(...)
}
So, to prevent this issue, we can slightly delay the destroy operation.
It is not entirely trivial, since we need to find a way to neutralize
the component in the meantime. But it seems like performing all that
kind of operation at the "commit" phase (so, the request animation frame
callback) makes sense to me.
So, this commit modifies the code to add a new component status
(cancelled) and use it to cancel components that are waiting to be
destroyed. These components will then be destroyed as soon as the
requestanimation frame starts, before all other dom operations.
Currently the implementation of batched is quite complicated and
difficult to read. This is because this approach tried to block all
calls at the same point and then only let the first one go through, but
an alternative approach is to simply throw away the calls that are made
after the first one has been scheduled. This change makes the
implementation much simpler to understand.
Co-authored-by: Aaron Bohy <aab@odoo.com>
Before this commit, owl parser would ignore the `.number` suffix on
<select> options. I do not see a good reason for that, and it prevents
some legitimate usecases.
closes#1444
This commit fixes a crash in the retrieval of the components tree which
happened when the first app did not contain a root node. The default
inspected component is now set to be the first root component found in
the apps.
This commit first fixes how object prototype are detected so that it
won't stop as soon as the constructor name of the object is "Object".
This allows displaying every single prototype encountered and closes
https://github.com/odoo/owl/issues/1467.
This commit also improves how the env of a component is displayed by
expanding its chain of prototypes by default while keeping its keys lit
as long as it is their first occurence in the chain.
This commit fixes the three following issues:
- Symbols could never appear in shortened display of objects
- Objects which contained only symbols as keys would be considered to be
empty and therefore not expandable
- Symbol value edition would create a new property on the object with
the stringified symbol as key instead of updating its value
closes https://github.com/odoo/owl/issues/1464
Since we now omit some observed objects when listing subscriptions, the
index of the subscription is no longer the index of the raw
subscription. This causes issue with the resolution of object paths when
they are inside a subscription. This commit fixes that by adding the
index to the raw subscription.
We also need to adapt the code that traverses the old tree, since the
index of the subscription in the component is no longer the same as the
index of the subscription "child" in the object tree.
Previously, the keys were displayed separately from the target, on top
of being displayed in bold inside the object when unfolded. This is
redundant, and in most cases you have to unfold the object anyway
because there are more keys observed than can be displayed.
The only "key" that cannot be displayed in bold inside the object is the
"key changes" magic key. This commit replaces it by a small +/- badge on
the right of the object's short content.
This commit also stops displaying observed object that are nested inside
other observed objects at the top level, as it can be confusing, and it
also heuristically gives a name to observed objects: if the observed
object is a property of the component or one of its props it will be
named accordingly, if not it will be named `[unknown]` (which may happen
when reobserving reactive objects inside a service or inside the env,
but is pretty rare in practice).
This commit changes the moment when all methods that need patching
to handle events are actually patched: the complete method of RootFiber
is now patched at devtools startup (when the tab owl tab is opened) and
every other method is patched at first activation of the event recording
functionality. This makes sure that the devtools will have no impact on
performance whatsoever when they are not directly put in use.
Class attributes are managed in a specific way in owl: they are merged
with existing class attributes, and also, they can be defined in
multiple ways (t-att-class, t-attf-class, t-att=['class', ...] and each
of these can be combined together.
Before this commit, the `t-att` syntax was handled as a normal
attribute, and therefore, would not combine as required with existing
classes.
closes#1453
This commit fixes some race conditions that were happening due to the
loading of the tree being sometimes triggered at the same moment as the
DOM patch. This would result in some destroyed component still being
present in the devtools' tree. This commit also ensures that the loading
of the tree is synchronous in regard to the component details loading.
Before this commit, the `t-on` directive assumed that the next character
would be a -, and ignored it, so if someone would write `t-onclick` by
mistake, Owl would then add an event handler on the `lick` event,
instead of `click`.
With this commit, we improve the parser to make it stricter and fail if
there is no dash.
closes#1441
This commit adds the functionnality to add components to the toggle
blacklist so that it won't be expanded by default when launching or
reloading the devtools components tree.
This commit adapts the behavior of the devtools global hook to get
the toRaw and reactive functions from the __OWL_DEVTOOLS__ variable
instead of the global owl when it is missing. Also adds the functions
to the __OWL_DEVTOOLS__ variable for the former to work.
This allows the devtools to work in environments where owl is loaded
through ES modules.
Also quickly fixes iframes detection update on page reload.
This commit fixes the key of the sub tree elements in the details
window to be 100% unique since the name of the elements could not
be unique so it would silently crash and not display. It also removes
the default size property of maps and sets since there's already a
getter available for it.
When refactoring the playground to use es-modules, a script that loads
owl as a non-module was forgotten in the head of the page and causes an
error in the console, as the script is not declared as type="module"
despite being an es-module. This script is also useless as owl is loaded
by being imported by the counter component in the page.
The importmap has also been moved to the head because of a bug in
firefox where importmaps can fail to be taken into account if they are
after a script that is not of type="module", see:
https://bugzilla.mozilla.org/show_bug.cgi?id=1833371Closes#1436
Before this commit, the translate function was not applied to text
content inside the body of a t-set (unless that content was itself
inside some html). This is not clearly not intended.
To fix it, we just need to call the translate function at the
appropriate moment.
closes#1434
Before this commit, when using a dynamic t-call, it was possible to have
a collision between subcomponent keys. This would cause hard to
understand owl crash, because owl would incorrectly use the same
component instance for two different template location. From
owl point of view, one of these component has to be destroyed and the
other one should be mounted, so a crash would occur.
The fix is to simply properly key sub components.
This commit improves how messages are relayed from the page to the
devtools by using a specific port for each instance of the owl devtools
panel and refreshing it through the heartbeat message. This allows
to prevent sending messages to every devtools instances and sorting
the messages inside.
This commit fixes how the apps methods are patched by decoupling
the patching of the destroy method from the rest because we need
to have a valid reference to a component node to do so (the root
node in this case). The previous behavior was to only patch methods
when an app with a root node is added but this was not working
when the app has no root node initially and gets one dynamically.
There is now a listener to patch the destroy method as soon as a
root node is set on any of the apps if it is not already the case.
# v2.1.2
- [FIX] t-out: does not crash when outputting null
- [FIX] playground: fix name of playground component
- [FIX] playground: use ES owl version instead of iife
- [IMP] playground: give a name so it is more explicit in devtools
- [REF] github page: move page to docs folder on master branch
- [IMP/FIX] playground: fix and improve various things
Before this commit, doing `t-out="null"` would crash, since the
safeOutput function would assume that it is a block, thanks to the fact
that `typeof null === 'object'`.
We handle here the null value just as if it was undefined: it outputs an
empty string.
Currently, some of the things on the playground must be changed directly
on the master branch while other things require checking out the
gh-pages branch and making the modifications there. Even when changes
need to be made on the master branch, all changes that are not direcly
in owl itself need to be copied by hand to the gh-pages branch.
This commit moves all files that exist on the gh-pages branch to the
master branch in the docs folder, this change will require configuring
the github page to use the docs folder instead of a separate branch, but
will make maintenance of the github page and playground easier going
forward.
The playground hasn't been touched for a while and a few things were
broken before this commit:
- Exporting as a standalone web application was broken because it still
attempted to load the templates like it was done in owl 1
- Resizing the editor tabs still tried to use `this.trigger`
While fixing the export feature, the files needed by the app were
factored out of hardcoded strings and into real files, as this makes it
easier to maintain, since these files get syntax highlighting.
The samples received the same treatment: there is now a `samples` folder
containing all the samples, it also contains a jsconfig, giving
autocompletion on owl functions while editing the files, and allowing
the owl import to look how it would when using owl as a node_module.
In the browser, this import is translated to the relative path of the
owl module using an import map.
# v2.1.1
- [FIX] components: solve missing update (concurrency issue)
- [DOC] remove old references to comp for useRef hook
- [FIX] devtools: Fix build commands for windows users
- [FIX] devtools: Fix devtools in detached window
- [IMP] devtools: Add a button to navigate to the doc
- [IMP] devtools: Add devtools documentation
- [IMP] Allow app to be mounted in shadow DOM
- [IMP] playground: allow sharing playground links
- [IMP] runtime: allow validating object values using a type description
- [FIX] devtools: Increase vertical padding of the search bar
- [FIX] devtools: Hide the collapse all button
- [FIX] devtools: Fix bad computation of highlight boxes on the page
- [FIX] package: Auto-update package-lock.json
Before this commit, in some specific situations, owl could skip updating
the DOM, even though it should. Here is what could happen:
- we have a parent component (A) and a child component (B)
- B depends on some reactive props from A
- we update some state that results in A being rendered, and B updated
- before this render is applied to the dom, we update some state again,
which causes A to be rendered again
- however, this new render is such that the B props are now identical,
so it will skip rendering B
- it will then apply the result of the render to the DOM => only A is
updated but B should also be updated!
The problem comes from the fact that Owl committed the props to the
component right after the rendering, to the new props were used in the
props comparison method to decide if B should be updated. This is
incorrect, we should always use the current props to decide what to do,
and only commit them to the component after it has been patched.
Going from owl 1.0 to 2.0, we removed the possibility of having a
reference to a component. However, the documentation was not properly
updated to reflect that.
This is useful to share editable snippets of owl code, which can be used
to paste a reproduction for a given issue or give someone an interactive
example of some feature.
In JS, objects are often used as a mapping that gives fast lookup on the
keys, in those cases, the keys themselves are not known in advance, but
the values may have a shape that they're expected to follow. This commit
adds support for a `values` key in schema for objects, which will be
checked against all values in the object during validation.
This commit adds box-sizing: border-box to the highlight elements
created by the devtools on the page. This was making some of the
boxes overflow on some pages (like on POS sessions) because
the border-box box-sizing was not the default behavior on these
pages.
Géry Debongnie