QWeb Templates

QWeb is the primary templating engine used by Odoo2. It is an XML templating engine1 and used mostly to generate HTML fragments and pages.

Template directives are specified as XML attributes prefixed with t-, for instance t-if for Conditionals, with elements and other attributes being rendered directly.

To avoid element rendering, a placeholder element <t> is also available, which executes its directive but doesn’t generate any output in and of itself:

<t t-if="condition">

will result in:


if condition is true, but:

<div t-if="condition">

will result in:


Data output

QWeb’s output directive out will automatically HTML-escape its input, limiting XSS risks when displaying user-provided content.

out takes an expression, evaluates it and injects the result in the document:

<p><t t-out="value"/></p>

rendered with the value value set to 42 yields:


See Advanced Output for more advanced topics (e.g. injecting raw HTML, etc…).


QWeb has a conditional directive if, which evaluates an expression given as attribute value:

    <t t-if="condition">

The element is rendered if the condition is true:


but if the condition is false it is removed from the result:


The conditional rendering applies to the bearer of the directive, which does not have to be <t>:

    <p t-if="condition">ok</p>

will give the same results as the previous example.

Extra conditional branching directives t-elif and t-else are also available:

    <p t-if="user.birthday == today()">Happy birthday!</p>
    <p t-elif="user.login == 'root'">Welcome master!</p>
    <p t-else="">Welcome!</p>


QWeb has an iteration directive foreach which take an expression returning the collection to iterate on, and a second parameter t-as providing the name to use for the “current item” of the iteration:

<t t-foreach="[1, 2, 3]" t-as="i">
    <p><t t-out="i"/></p>

will be rendered as:


Like conditions, foreach applies to the element bearing the directive’s attribute, and

<p t-foreach="[1, 2, 3]" t-as="i">
    <t t-out="i"/>

is equivalent to the previous example.

foreach can iterate on an array (the current item will be the current value) or a mapping (the current item will be the current key). Iterating on an integer (equivalent to iterating on an array between 0 inclusive and the provided integer exclusive) is still supported but deprecated.

In addition to the name passed via t-as, foreach provides a few other variables for various data points:


$as will be replaced by the name passed to t-as

$as_all (deprecated)

the object being iterated over


This variable is only available on JavaScript QWeb, not Python.


the current iteration value, identical to $as for lists and integers, but for mappings it provides the value (where $as provides the key)


the current iteration index (the first item of the iteration has index 0)


the size of the collection if it is available


whether the current item is the first of the iteration (equivalent to $as_index == 0)


whether the current item is the last of the iteration (equivalent to $as_index + 1 == $as_size), requires the iteratee’s size be available

$as_parity (deprecated)

either "even" or "odd", the parity of the current iteration round

$as_even (deprecated)

a boolean flag indicating that the current iteration round is on an even index

$as_odd (deprecated)

a boolean flag indicating that the current iteration round is on an odd index

These extra variables provided and all new variables created into the foreach are only available in the scope of the foreach. If the variable exists outside the context of the foreach, the value is copied at the end of the foreach into the global context.

<t t-set="existing_variable" t-value="False"/>
<!-- existing_variable now False -->

<p t-foreach="[1, 2, 3]" t-as="i">
    <t t-set="existing_variable" t-value="True"/>
    <t t-set="new_variable" t-value="True"/>
    <!-- existing_variable and new_variable now True -->

<!-- existing_variable always True -->
<!-- new_variable undefined -->


QWeb can compute attributes on-the-fly and set the result of the computation on the output node. This is done via the t-att (attribute) directive which exists in 3 different forms:


an attribute called $name is created, the attribute value is evaluated and the result is set as the attribute’s value:

<div t-att-a="42"/>

will be rendered as:

<div a="42"></div>

same as previous, but the parameter is a format string instead of just an expression, often useful to mix literal and non-literal string (e.g. classes):

<t t-foreach="[1, 2, 3]" t-as="item">
    <li t-attf-class="row {{ (item_index % 2 === 0) ? 'even' : 'odd' }}">
        <t t-out="item"/>

will be rendered as:

<li class="row even">1</li>
<li class="row odd">2</li>
<li class="row even">3</li>


There are two equivalent syntaxes for format strings: "plain_text {{code}}" (aka jinja-style) and "plain_text #{code}" (aka ruby-style).


if the parameter is a mapping, each (key, value) pair generates a new attribute and its value:

<div t-att="{'a': 1, 'b': 2}"/>

will be rendered as:

<div a="1" b="2"></div>

if the parameter is a pair (tuple or array of 2 element), the first item of the pair is the name of the attribute and the second item is the value:

<div t-att="['a', 'b']"/>

will be rendered as:

<div a="b"></div>

setting variables

QWeb allows creating variables from within the template, to memoize a computation (to use it multiple times), give a piece of data a clearer name, …

This is done via the set directive, which takes the name of the variable to create. The value to set can be provided in two ways:

  • a t-value attribute containing an expression, and the result of its evaluation will be set:

    <t t-set="foo" t-value="2 + 1"/>
    <t t-out="foo"/>

    will print 3

  • if there is no t-value attribute, the node’s body is rendered and set as the variable’s value:

    <t t-set="foo">
    <t t-out="foo"/>

calling sub-templates

QWeb templates can be used for top-level rendering, but they can also be used from within another template (to avoid duplication or give names to parts of templates) using the t-call directive:

<t t-call="other-template"/>

This calls the named template with the execution context of the parent, if other_template is defined as:

<p><t t-value="var"/></p>

the call above will be rendered as <p/> (no content), but:

<t t-set="var" t-value="1"/>
<t t-call="other-template"/>

will be rendered as <p>1</p>.

However, this has the problem of being visible from outside the t-call. Alternatively, content set in the body of the call directive will be evaluated before calling the sub-template, and can alter a local context:

<t t-call="other-template">
    <t t-set="var" t-value="1"/>
<!-- "var" does not exist here -->

The body of the call directive can be arbitrarily complex (not just set directives), and its rendered form will be available within the called template as a magical 0 variable:

    This template was called with content:
    <t t-out="0"/>

being called thus:

<t t-call="other-template">

will result in:

    This template was called with content:

Advanced Output

By default, out should HTML-escape content which needs to be escaped, protecting the system against XSS

Content which does not need to be escaped will instead be injected as-is in the document, and may become part of the document’s actual markup.

The only cross-platform “safe” content is the output of t-call or a t-set used with a “body” (as opposed to t-value or t-valuef).


Usually you should not have to care too much: APIs for which it makes sense should generate “safe” content automatically, and things should work transparently.

For the cases where things need to be clearer though the following APIs output safe content which will by default not be (re-)escaped when injected into templates:

  • HTML fields.

  • html_escape() and markupsafe.escape() (they are aliases, and have no risk of double-escaping).

  • html_sanitize().

  • markupsafe.Markup.


    markupsafe.Markup is an unsafe API, it’s an assertion that you want the content to be markup-safe but necessarily can not check that, it should be used with care.

  • to_text() does not mark the content as safe, but will not strip that information from safe content.

Creating safe content using Markup

See the official documentation for explanations, but the big advantage of Markup is that it’s a very rich type overriding str operations to automatically escape parameters.

This means that it’s easy to create safe html snippets by using Markup on a string literal and “formatting in” user-provided (and thus potentially unsafe) content:

>>> Markup('<em>Hello</em> ') + '<foo>'
Markup('<em>Hello</em> &lt;foo&gt;')
>>> Markup('<em>Hello</em> %s') % '<foo>'
Markup('<em>Hello</em> &lt;foo&gt;')

though it is a very good thing, note that the effects can be odd at times:

>>> Markup('<a>').replace('>', 'x')
>>> Markup('<a>').replace(Markup('>'), 'x')
>>> Markup('<a&gt;').replace('>', 'x')
>>> Markup('<a&gt;').replace('>', '&')


Most of the content-safe APIs actually return a Markup with all that implies.



Due to the lack of operator overriding, Markup() is a much more limited type than Markup.

Therefore it doesn’t override methods either, and any operation involving Markup() will return a normal String() (and in reality not even that, but a “primitive string”).

This means the fallback is safe, but it is easy to trigger double-escaping when working with Markup() objects.

forcing double-escaping

If content is marked as safe but for some reason needs to be escaped anyway (e.g. printing the markup of an HTML fields), it can just be converted back to a normal string to “strip” the safety flag e.g. str(content) in Python and String(content) in Javascript.


Because Markup is a much richer type than Markup(), some operations will strip the safety information from a Markup() but not a Markup e.g. string concatenation ('' + content) in Python will result in a Markup with the other operand having been properly escaped, while in Javascript will yield a String() where the other operand was not escaped before the concatenation.

Deprecated output directives


An alias for out, would originally HTML-escape its input. Not yet formally deprecated as the only difference between out and esc is that the latter is a bit unclear / incorrect.


A version of out which never escapes its content. Content is emitted as-is, whether it’s safe or not.

Deprecated since version 15.0: Use out with a markupsafe.Markup value instead.

t-raw was deprecated because as the code producting the content evolves it can be hard to track that it’s going to be used for markup, leading to more complicated reviews and more dangerous lapses.


Exclusive directives

Asset bundles

“smart records” fields formatting

The t-field directive can only be used when performing field access (a.b) on a “smart” record (result of the browse method). It is able to automatically format based on field type, and is integrated in the website’s rich text editing.

t-options can be used to customize fields, the most common option is widget, other options are field- or widget-dependent.



invokes a debugger using PDB’s set_trace API. The parameter should be the name of a module, on which a set_trace method is called:

<t t-debug="pdb"/>

is equivalent to importlib.import_module("pdb").set_trace()

Rendering cache:

t-cache="key_cache" tags part of template to be cached at rendering time. Every sub-directives will be call only during the first rendering. It means that the sql queries excecuted during the rendering of those sub-directives are also done only once.

t-nocache="documentation" tags part of template to be render every time. The content can only use the root values.

Why and when to use t-cache?

This directive is used to speed up the rendering, by caching parts of the final document, which may save queries to the database. However, it should be used sparingly, as t-cache will inevitably complicate templates (and their understanding of t-set for example).

However, in order to actually save database queries, it might be necessary to render the template with values that are evaluated lazily. If those lazy values are used in a cached part, they will not be evaluated if the part is available in cache.

The t-cache directive is useful for template parts using values that depend on a limited amount of data. We recommend to analyze the rendering of the template with the profiler (by activating the “Add qweb directive context” option). Passing lazy values to the rendering in controllers allow you to display the directives using these values and triggering the queries.

A concern of using such a cache are making it available to different users (different users should render the cached parts the same way). Another potential issue is to invalidate its entries when necessary. For the latter, the key expression should be chosen wisely. Using the write_date of a recordset can make a cache key out-of-date without having to discard it explicitly from the cache, for instance.

One should also pay attention to the fact that the values in t-cache parts are scoped. This implies that if there are t-set directives in this part of the template, the rendering of what comes after it could be different than if there was no t-cache directive.

What if there is a t-cache inside a t-cache?

The parts are cached. Each containing only the string corresponding to its rendering. Thus, the t-cache inside will probably be read less often, its cache key will not necessarily be used. If this must be the case, then you may need to add a t-nocache (on the same node or a parent).

What is t-nocache used for?

If you want to cache part of a template with t-cache but a small piece must remain dynamic and be evaluated at cache times. However, the part in t-nocache will not have access to the t-set value of the template. Only the values ​​provided by the controller are accessible there. For example, the menu is cached because it’s the same all the time and takes time to render (using the performance devtools with the qweb context lets you investigate). However, in the menu, we want the ecommerce cart to be always up to date. So there is a t-nocache to keep this part dynamic.

The base of t-cache

The t-cache directive allows you to store the rendered result of a template. The key expression (eg 42: t-cache="42") will be evaluated as a python expression. This will be used to generate the cache key. So there can be different cache values (cached render part) for the same template part. If the key expression is a tuple or a list, it will be searched when generating the cache key. If one or more recordsets are returned by the key expression, then the model, ids and their corresponding write_date will be used to generate the cache key. Special case: If the key expression returns a Falsy value, then the content will not be cached.


<div t-cache="record,bool(condition)">
    <span t-if="condition" t-field="record.partner_id.name">
    <span t-else="" t-field="record.partner_id" t-options-widget="contact">

In this case, there may be values ​​(string) in the cache corresponding to each record already returned with a true condition, as well as for the false condition. And if a module modifies the record, the write_date being modified, the cached value is discarded.

t-cache and scoped values (t-set, t-foreach…)

Values in t-cache are scoped, this involves a change in behavior between having or not having t-cache on one of the parent nodes. Don’t forget to take into account that Odoo uses a lot of templates, t-call and view inheritance. Adding a t-cache can therefore modify the rendering of a template that you do not see when editing. (t-foreach it’s like a t-set for each iteration)


    <t t-set="a" t-value="1"/>
        <t t-set="a" t-value="2"/>
        <t t-out="a"/>
    <outside t-out="a"/>

    <t t-set="b" t-value="1"/>
    <inside t-cache="True">
        <t t-set="b" t-value="2"/>
        <t t-out="b"/>
    <outside t-out="b"/>

Will render:



The base of t-nocache

The template part contained in a node with a t-nocache attribute is not cached. This content is therefore dynamic and is rendered systematically. However the available values are those provided by the controller (when calling the _render method).


    <article t-cache="record">
        <title><t t-out="record.name"/> <i t-nocache="">(views: <t t-out="counter"/>)</i></titlle>
        <content t-out="record.description"/>

Will render (counter = 1):

        <title>The record name <i>(views: 1)</i></titlle>
        <content>Record description</content>

Here the <i> tag that contains the container will always be rendered. While the rest is as a single string in the cache.

t-nocache and scoped root values (t-set, t-foreach…)

The contents of the t-nocache tag can be used for documentation and to explain why the directive is added. The values are scoped into t-nocache, these values are root values only (values provided by the controller and/or when calling the _render method of ir.qweb). t-set can be done in the template part, but will not be available elsewhere.


    <t t-set="counter" t-value="counter * 10"/>
    <header t-nocache="">
        <t t-set="counter" t-value="counter + 5"/>
        (views: <t t-out="counter"/>)
    <article t-cache="record">
        <title><t t-out="record.name"/> <i t-nocache="">(views: <t t-out="counter"/>)</i></titlle>
        <content t-out="record.description"/>
    <footer>(views: <t t-out="counter"/>)</footer>

Will render (counter = 1):

        (views: 6)
        <title>The record name <i>(views: 1)</i></titlle>
        <content>Record description</content>
    <footer>(views: 10)</footer>

Here the <i> tag that contains the container will always be rendered. While the rest is as a single string in the cache. The counter is not updated by the t-set out of the t-nocache

t-nocache-* add some primitive values in the cache

In order to be able to use values generated in the template, it is possible to cache them. The directive is used as t-nocache-*="expr" where * is the name of the chosen value and expr the python expression so the result will be cached. The cached value must be primitive type.


<section t-cache="records">
    <article t-foreach="records" t-as="record">
            <title t-field="record.get_method_title()"/>
        <footer t-nocache="This part has a dynamic counter and must be rendered all the time."
            <span t-out="counter + cached_value"/>

The value cached_value is cached with the cached template part of t-cache="records" and add to the scoped root values each time.



Most Python-side uses of QWeb are in controllers (and during HTTP requests), in which case templates stored in the database (as views) can be trivially rendered by calling odoo.http.HttpRequest.render():

response = http.request.render('my-template', {
    'context_value': 42

This automatically creates a Response object which can be returned from the controller (or further customized to suit).


At a deeper level than the previous helper is the _render method on ir.qweb (use the datable) and the public module method render (don’t use the database):

_render(id[, values])

Renders a QWeb view/template by database id or external id. Templates are automatically loaded from ir.qweb records.

_prepare_environment method sets up a number of default values in the rendering context. The http_routing and website addons, also default values they need. You can use minimal_qcontext=False option to avoid this default value like the public method render:


the current Request object, if any


whether the current request (if any) is in debug mode


url-encoding utility function


the corresponding standard library module


the corresponding standard library module


the corresponding standard library module


see module


the keep_query helper function

  • values – context values to pass to QWeb for rendering

  • engine (str) – name of the Odoo model to use for rendering, can be used to expand or customize QWeb locally (by creating a “new” qweb based on ir.qweb with alterations)

render(template_name, values, load, **options)

returns etree object, ref


Exclusive directives

Defining templates

The t-name directive can only be placed at the top-level of a template file (direct children to the document root):

    <t t-name="template-name">
        <!-- template code -->

It takes no other parameter, but can be used with a <t> element or any other. With a <t> element, the <t> should have a single child.

The template name is an arbitrary string, although when multiple templates are related (e.g. called sub-templates) it is customary to use dot-separated names to indicate hierarchical relationships.

Template inheritance

Template inheritance is used to either:
  • Alter existing templates in-place, e.g. to add information to templates

created by other modules.
  • Create a new template from a given parent template

Template inheritance is performed via the use of two directives:
  • t-inherit which is the name of the template to inherit from,

  • t-inherit-mode which is the behaviour of the inheritance: it can either be set to primary to create a new child template from the parented one or to extension to alter the parent template in place.

An optional t-name directive can also be specified. It will be the name of the newly created template if used in primary mode, else it will be added as a comment on the transformed template to help retrace inheritances.

For the inheritance itself, the changes are done using xpaths directives. See the XPATH documentation for the complete set of available instructions.

Primary inheritance (child template):

<t t-name="child.template" t-inherit="base.template" t-inherit-mode="primary">
    <xpath expr="//ul" position="inside">
        <li>new element</li>

Extension inheritance (in-place transformation):

<t t-inherit="base.template" t-inherit-mode="extension">
    <xpath expr="//tr[1]" position="after">
        <tr><td>new cell</td></tr>

Old inheritance mechanism (deprecated)

Template inheritance is performed via the t-extend directive which takes the name of the template to alter as parameter.

The directive t-extend will act as a primary inheritance when combined with t-name and as an extension one when used alone.

In both cases the alteration is then performed with any number of t-jquery sub-directives:

<t t-extend="base.template">
    <t t-jquery="ul" t-operation="append">
        <li>new element</li>

The t-jquery directives takes a CSS selector. This selector is used on the extended template to select context nodes to which the specified t-operation is applied:


the node’s body is appended at the end of the context node (after the context node’s last child)


the node’s body is prepended to the context node (inserted before the context node’s first child)


the node’s body is inserted right before the context node


the node’s body is inserted right after the context node


the node’s body replaces the context node’s children


the node’s body is used to replace the context node itself


the nodes’s body should be any number of attribute elements, each with a name attribute and some textual content, the named attribute of the context node will be set to the specified value (either replaced if it already existed or added if not)

No operation

if no t-operation is specified, the template body is interpreted as javascript code and executed with the context node as this


while much more powerful than other operations, this mode is also much harder to debug and maintain, it is recommended to avoid it


The javascript QWeb implementation provides a few debugging hooks:


takes an expression parameter, evaluates the expression during rendering and logs its result with console.log:

<t t-set="foo" t-value="42"/>
<t t-log="foo"/>

will print 42 to the console


triggers a debugger breakpoint during template rendering:

<t t-if="a_test">
    <t t-debug=""/>

will stop execution if debugging is active (exact condition depend on the browser and its development tools)


the node’s body is javascript code executed during template rendering. Takes a context parameter, which is the name under which the rendering context will be available in the t-js’s body:

<t t-set="foo" t-value="42"/>
<t t-js="ctx">
    console.log("Foo is", ctx.foo);



(core is the web.core module) An instance of QWeb2.Engine() with all module-defined template files loaded, and references to standard helper objects _ (underscore), _t (translation function) and JSON.

core.qweb.render can be used to easily render basic module templates


class QWeb2.Engine()

The QWeb “renderer”, handles most of QWeb’s logic (loading, parsing, compiling and rendering templates).

Odoo Web instantiates one for the user in the core module, and exports it to core.qweb. It also loads all the template files of the various modules into that QWeb instance.

A QWeb2.Engine() also serves as a “template namespace”.

QWeb2.Engine.QWeb2.Engine.render(template[, context])

Renders a previously loaded template to a String, using context (if provided) to find the variables accessed during template rendering (e.g. strings to display).

  • template (String()) – the name of the template to render

  • context (Object()) – the basic namespace to use for template rendering



The engine exposes an other method which may be useful in some cases (e.g. if you need a separate template namespace with, in Odoo Web, Kanban views get their own QWeb2.Engine() instance so their templates don’t collide with more general “module” templates):


Loads a template file (a collection of templates) in the QWeb instance. The templates can be specified as:

An XML string

QWeb will attempt to parse it to an XML document then load it.


QWeb will attempt to download the URL content, then load the resulting XML string.

A Document or Node

QWeb will traverse the first level of the document (the child nodes of the provided root) and load any named template or template override.

A QWeb2.Engine() also exposes various attributes for behavior customization:


Prefix used to recognize directives during parsing. A string. By default, t.


Boolean flag putting the engine in “debug mode”. Normally, QWeb intercepts any error raised during template execution. In debug mode, it leaves all exceptions go through without intercepting them.


The jQuery instance used during template inheritance processing. Defaults to window.jQuery.


A Function. If present, called before compiling each DOM node to template code. In Odoo Web, this is used to automatically translate text content and some attributes in templates. Defaults to null.


it is similar in that to Genshi, although it does not use (and has no support for) XML namespaces


although it uses a few others, either for historical reasons or because they remain better fits for the use case. Odoo 9.0 still depends on Jinja and Mako.