With multi-websites, we expect qweb views to have a key set as it is the key
that is used to find duplicates (two views with same keys are duplicate, the
one with a website_id set is more specific than the one without a website_id)
Thus, 5ff87e8039 sort on key in order to find the most suitable one.
It would crash if a qweb view is created without a key (False) as it can't sort
`bool` and `str`.
This commit:
- Adds a generated key to scss views
- Adds an SQL constraint to avoid QWeb views without a `key`
- Generates a random key when creating a QWeb view without a `key`
- Handle False key in `filter_duplicate()` by extracting views with False key
before sorting, and then adding these views to the recordset
Note: We also want the key to be editable as it is now an important field with
multiwebsite. Thus, we removed the readonly on this field.
+ fix python tests as we now force key on qweb views with sql constraint
+ pep8
Fix bug from the initial poc of JOV
Fix menu creation
- creating a menu would create a 'container' menu in the DB (.create is call without website_id)
then writing on it would copy (if condition to cow) the menu with a website_id leaving the first
one as a menu container unused.
- website_menu should always have a website_id
we dont want to support the multiwebsite system that allow to have generic/specific menu
unlink/write is useless now since we always have a website_id
Make unique_path website dependent, 2 distinct website can have a page with same name
Make Sale report multiwebsite compliant
Make website_id on sale_order / account.invoice as a related stored from partner_id.
Co-authored-by: Derie Romain <rde@odoo.com>
Co-authored-by: Kersten Jérémy <jke@odoo.com>
This implements support to administer multiple websites. Although the
core functionality already existed, managing multiple websites was
fairly technical.
In the interest of database updates and migration this attempts to
keep duplicated data to a minimum. To do this the usual generic
records are rendered unless some website-specific record exists that
replaces it. Copy-on-write (COW) is used to create these
website-specific records. Through this mechanism creating a
website-specific record is delayed until necessary. A COW mechanism
has been implemented on 4 models: ir.ui.view, website.page,
website.menu and ir.attachment. These COW mechanisms are activated
when editing data through the website (aka frontend). These frontend
edits (e.g. with web_editor) will be website-specific, possibly
creating a website-specific record when necessary. When editing data
in the backend nothing special will happen, even when editing a
generic record. Note that because of this mechanism also facilitates
the ability to create new, uncustomized websites because the generic
data is kept.
Support is provided for a website to have any theme. Themes are fairly
complex to handle. Standalone themes can depend on other standalone
themes (e.g. theme_beauty depends on theme_loftspace) and themes
usually modify some data of the themes they depend on. Because a theme
can be installed on multiple websites, using website_id m2o fields
does not work well. It would require duplicate data, making updates
and migration harder. Because of this, data for themes (ir.ui.view and
ir.attachment specifically) have a theme_id m2o. website has a
theme_ids m2m that identifies all theme modules currently installed on
it. Through these fields we figure out what to render. A theme is only
fully uninstalled when it's no longer active on any website. The
advantage of this approach is that upgrading or migrating theme data
is no different from the single-website case.
The website.published.mixin class was modified to handle multiple
websites. A wizard was added in the backend to easily manage this for
multiple website.
Although not used anywhere in this commit, a 'website_id' variable has
been added in the evaluation context of ir.rule. It allows to easily
make any model multi-website aware, all that's needed is a custom
website_id m2o field on a model and a custom record rule.
Refactoring of qweb to call _post_processing_att for each nodes. This
change remove the crappy ovewrite in website module. The website overwrite
only the _post_processing_att to add the cdn parameters.
The static node (without t- attributes) can stay static (remove overwrite
of _is_static_node), the cdn is applied at the compile time for this
nodes instead of at the running time like the dynamic node.
Purpose: When running tests, all the tests for the installed/updated
files are done. This commit adds a 'tagged' decorator that can be used to
tag tests. Combined with a new 'test-tags' CLI option, it adds the ability
to filter which tests are executed. For example, @tagged('slow') will
add a tag 'slow' to the test. The CLI option 'test-tags="slow"' will
only run tests tagged 'slow'.
One can use prefixes to select cases with tags.
'+' or no prefix means that the tests tagged with this tag are selected
for execution. '-' prefix will exclude the tests tagged with this tag.
Exclusion takes precedence over inclusion.
Also, by default, all Odoo tests cases are tagged 'standard' and with
the technical name of the module.
This means that when selecting tests with the 'test-tags'
parameter, if '-standard' is not specified, all tests tags are
going to be executed.
When tagging tests, one can remove such automatic tag by prefixing the
tag name with '-'. E.g. @tagged('-standard') will remove the standard
tag from the test.
Another example, if one wants to test the 'sale' module alone,
even without adding any 'tagged' decorator thos tests can be selected
like that: --test-tags="sale"
Tests are selected or deselected using a TagsSelector. When instanciated,
a string is passed with comma separated tests selectors like
'+slow,-standard'. When the 'check' method is called with a test as argument,
it returns True or False if the test has to be executed or not.
Now that we're closer to switching to P3 for good, these helpers have
outlived their usefulness, and mostly add noise.
All remaining dict.iter*() or dict.view*() must be converted to the
normal keys(), values() or items() calls.
Whenever the result is likely to be used for more than the scope of a
loop, or when the dict needs to be modified during iteration, the calls
must be wrapped in a ``list()``, to protect the new P3 semantics.
Those cases are very exceptional.
Also removed some dead code or improved the API to remove unnecessary
conversions.
* mass_mailing, payment, point_of_sale, portal, survey, web, web_tour,
website_blog, website_crm_partner_assign, website_event,
website_event_questions, website_form, website_forum, website_gengo,
website_hr_recruitment, website_links, website_mail,
website_mail_channel, website_mass_mailing, website_quote,
website_sale, website_sale_options, website_slides, website_twitter
This commit reviews the whole "JS side" of the web_editor and website
apps. This is a first step to be able to improve them with new and
better functionnalities; this commit is not supposed to change any
visual behavior.
The main goal was to achieve a structure similar to the backend one.
Now, the frontend side also has a root widget (like the WebClient)
and all other widgets are attached to it one way or another. This allows
the benefits of using the 'trigger_up' functionnality for example.
As RPC are now mainly done with the `this._rpc` functionnality (being
possible thanks to the parent hierarchy), the frontend will also be
possible to test thanks to QUnit in a future update (besides the "text"
editor side which still requires a refactoring to be able to do that).
---
Here are some of the changes:
(-) conventions and documentation
The code has been updated to follow JS conventions and a lot of code has
been commented (around +2000 lines of comment). This also means that
lots of functions have been renamed to use camelCase or simply to make
their name understandable.
See https://github.com/odoo/odoo/wiki/Javascript-coding-guidelines.
(-) deprecated: web_editor.base
The "web_editor.base" module has been split and does not force the
modules which require it to wait for DOM ready anymore. This was indeed
slowing loading times, but also prevented to use some modules in some
contexts (see the LESS editor use in web_studio which is the subject of
another task).
Now the "editor context" can be got thanks to the "web_editor.context"
JS module with its "get" function.
The "web_editor.base" module should probably not be used anymore (see
its code and recent updates).
(-) new: web.dom_ready
If a JS module should wait for the DOM to be ready to be executed, a
new JS module has been created: "web.dom_ready". This should always
be used in a module which only want to instantiate stuff. Do not
extend (or worst, include) classes after DOM ready.
(-) website.website
The "website.website" module has been split. "website.website" does not
return anything useful anymore, it just initialize some miscellaneous
stuff, without waiting for the DOM to be ready. You might want to check
"website.utils", "website.content.compatibility" or `WebsiteRoot`. Also
`website.form` has been deleted (use `this._rpc`), so has been
`website.error`. `website.prompt` will be removed in a future update to
be replaced by `Dialog.prompt`.
(-) widgets are great
Many classes which were not widgets are now widgets. This allows them to
use the 'events', the 'xmlDependencies' and the 'this._rpc' features for
example. Here are some of the main ones:
- Snippet options: these were classes with a `$el` for the menu element
and `$target` for the customized element. This is still the case
but following standard `Widget` structure (one exception: using
`this.$(...)` searches in the `$target` as before this update).
- Snippet animations: instead of class instances with a `$target`
element which can be `start` and `stop`, these are now standard
widgets which can be `start` and `destroy`. `this.$target` is
an alias to `this.$el` for ease of compatibility.
- Snippet editors: instead of class instances in charge of an editor
overlay, these are now widgets. Each "child" snippet editor is
properly attached as a "child", which allows editors to communicate
and to be properly destroyed.
(-) root widgets and website navbar
The frontend is different of the backend. In the backend, the page has
an empty <body/> element and all the components are instantiated from
parent to children (i.e. the `WebClient` is instantiated and is in
charge of instantiating the `ControlPanel`, etc). The frontend cannot
work like that on page loadings as they are way more frequent than in
the backend and we do not want them to flicker. A frontend page is
loaded as a <body/> element which already contains the website navbar
and its menus and the whole content page. JS code has to be "attached"
to these existing elements. This is possible thanks to the `RootWidget`
instances and the specialized `WebsiteRoot`, `IframeRoot` and
`WebsiteNavbar` (see code for details).
(-) lazy loading
No more (or at least a lot less) XML/JS has to be loaded on page
loading, thanks to the use of the `Widget.xmlDependencies` feature.
XML which have to be lazy loaded is loaded only on related Widget
instantiation if necessary, which allows to execute a lot of JS code
before the DOM is ready and to start many widgets on DOM ready (not
later). A visual benefit of this is clicking on the 'edit' button as
soon as it is possible: before this commit, this was sometimes not
doing anything as event handlers were not binded yet.
Still a possible exception: loading the session and locales. This may
be asynchronous stuff which is still required before widget
instantiations but this will be the subject of another task.
(-) deprecated code and code location
More than reviewing code and organizing it, many apparent dead code was
removed. More importantly, mislocated code was put in the right app.
This is the case for snippet animations which is a concept for website
apps but was defined in the web_editor app, or some translation concepts
which were part of website but should have been part of web_editor.
---
There are probably more things to say about this commit but I will let
the comments speak for those.
Slug and unslug API is now available in http_routing. Indeed there is no
link to any website or any reason to not support slufigied URLs when
website is not installed. A new unslug_url method is added as a tool
coming from an embedded method from website. Doing it allows to have all
slug related methods defined at the same point.
Support of slug and unslug in qweb rendering is also moved directly in
http_routing version of ir_ui_view instead of website inheritance. This
is done in order to keep things coherent.
Some code from website about ModelConverter is also moved. Indeed both
versions of ir_http uses some kind of placeholder to store the uid
when converting urls to python. This commit unifies it by using the
website one directly in base to simplify the override.
This commit also updates all module importing slug. Enterprise modules
will have to be updated, see related commit.
In Python 3, all of these were "consolidated" under urllib(.request,
.parse, .errors) which is inconvenient.
Since we already have hard dependencies on requests and
werkzeug(.urls, which is a backport of Python 3's unicode-aware
urllib.parse) migrate *everything* to that.
A sticking point is urllib2.URLError, those were (mostly) replaced by
the slightly more general IOError which URLError extends.
In Python 3:
* various builtins and dict methods were changed to return
view/iterable objects rather than lists
* and the separate Python 2 view/iterable builtins and methods were
removed altogether
This is problematic when using these items as list (which the happens
repeatedly in Odoo), but more viciously when iterating *multiple times*
over them (which also happens, which I've messed up multiple times while
writing this, and which is a pain to debug even when you've just created
the issue).
Convert all code using these to semantics-matching cross-version
helper functions to get the LCD behaviour between P2 and P3, and
forbid the builtins via lint.
issue #8530
QWeb's get_template:
* remove the fallback that returned the entire result of IRQweb's
`load` if no match was made between the t-name of the first children
of this result and the template to render
* When get_template receives an etree, directly return it and don't
try to find a matching t-name in it. The loader was not called but
a matching was trying to be made, now that we removed the fallback
a lot of tests that were relying on passing etree didn't work
anymore.
IRQweb's load now return consistently the full result of IrUiView's
`read_templace` when asked to load an id or an xmlid. We make a special
case to patch the result if asked with an id or with a child view so
that the "t-name" of one of the first children of what we return is
always what QWeb is looking for.
This fixes the problem of recomputing that field in a much better way. The
field was stored for the sake of searching; instead we implemented a search
method for it.
The field was a stored computed field without dependencies, and therefore it
was never (re)computed. Add a simple dependency on the model `ir.model.data`
to make the recomputation everytime a view is modified. Moreover, when views
are created from modules, delay the recomputation of view fields until their
`ir.model.data` record is created.
* crm, project, website, website_event, website_blog, website_forum,
website_sale
Eg: the tour 'shop_buy_product' is extended by the website_sale_options
addons to add a step to close a modal.
The current solution was requiring the module that defines the tour
to extend, then add/remove steps in the "step" key of the tour
definition.
Some of the problems with this method were:
* The "register" method calls the "update" method to immediately
search for tip to place once registered (as register may be called
after the DOM is ready). So extensions of tours were happening after
the tours were started (and the tours were not restarted).
* The addition/removal of steps was happening after they were filtered
according to the "edition" key.
To allow extension, the system is changed as follow:
* The "register" method now only saves the steps and options without
modifying them and does not call the "update" method.
* Once the DOM was ready, the tour service started listening to DOM
mutations and called the "update" tour method. Now, this "update" call
is replaced by a "_register_all" call, on DOM ready and at the end of
the current call stack (which makes sure all modules are loaded). This
"_register_all" method marks the registered tours as ready after having
filtered the steps according to the "edition" key and initialized the
current step to trigger.
* Also, tours can now define a "wait_for" option which allow them to
be marked as ready for run and update after the given deferred. Those
tours can now also be extended without having to wait for the deferred.
PhamtomJS must wait for the "ready" key to be true to run the tour.
Since saas-3, website controllers use `request.website.render`
to render the template of a web page. This was kept for retro
compatibility. It's time to stop using deprecated stuff.
Same for `_render` method on website.
'json' route don't use `request.render` since
`JsonRequest` has no `render` method.
Migration of website module to new API.
The tricky phase is ir_http.py : we need to use `request.env`
only when the authenfication phase is done. Normally by
calling `super` of `_dispatch` method, but website module
required it to be done before. This is important since
`env`is a lazy property of `request` object.
Some hack were kept since this commit is a migration ('RequestUID'
in ir.http, ...)
Some docstrings were added.
The `unittest2` package is simply a backport of `unittest` from the
Standard Library of Python 2.7 to previous versions.
There is no reason to use it any longer.
Closes#6941
Escape text nodes changed via the web editor before sending the content
it to the server controller.
It is done since the content is unescaped one time when being displayed,
and it is not done for inline style and script tags (which may be
injected by dropping a snippet) since that would break them.
replacing the solution in cdb900044.
When saving a template in version 8.0, html would be saved as it should
be displayed once on the site. In particular, if some text should be
escaped once send to the browser, it will be saved as such.
But when rendering, a text node content is unescaped two times:
* for translation which seems wrong since we already use .text of a node
which already escaped it, doing it one more time is bad,
* when rendering the template, since the html template is stored in xml,
This commit remove superfluous unescaping for translation, and add an
escaping when saving the changed template content.
closes#7967
opw-646889
If they are this routes:
/partner/p-1
/partner/p-2
...
/partner/grade-1/p-1
/partner/grade-1/p-2
...
/partner/grade-2/p-1
/partner/grade-2/p-2
...
We want test only one time the routes:
/partner/p-1
/partner/grade-1/p-1
The old-api model._all_columns contains information about model._columns and
inherited columns. This dictionary is missing new-api computed non-stored
fields, and the new field objects provide a more readable api...
This commit contains the following changes:
- adapt several methods of BaseModel to use fields instead of columns and
_all_columns
- copy all semantic-free attributes of related fields from their source
- add attribute 'group_operator' on integer and float fields
- base, base_action_rule, crm, edi, hr, mail, mass_mailing, pad,
payment_acquirer, share, website, website_crm, website_mail: simply use
_fields instead of _all_columns
- base, decimal_precision, website: adapt qweb rendering methods to use fields
instead of columns
Romain Derie