Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Introduce a CorrectnessNodeVisitor to validate that templates are semantically correct #4292

Open
wants to merge 1 commit into
base: 3.x
Choose a base branch
from

Conversation

fabpot
Copy link
Contributor

@fabpot fabpot commented Sep 9, 2024

This PR addresses several issues around the correctness of templates.
Being able to parse and compile a template does not mean that it is semantically correct. To enforce correctness, we currently have several places where we deal with it:

  • Parser::filterBodyNodes(): This method is a mix of ensuring the correctness of a template, but it also changes the body node of a child template (something that is always needed and not part of the correctness checks)
  • ExtendsTokenParser: It checks that an extend tag is not embedded into a block or a macro. The extend tag is not the only one that must be at the root of a template

This PR introduces a new CorrectnessNodeVisitor that has the responsibility to check that a template is semantically correct. It's the continuation of work that started a long time ago in #2687 (where I mentioned the weirdness of some supported templates like those mentioned in #3926 and deprecated by this PR).

Closes #3698: Having a use tag embedded in another tag (like if in the mentioned PR) is deprecated and will not be possible in 4.0.


$body = $node->getNode('body')->getNode('0');
// see Parser::subparse() which does not wrap the parsed Nodes if there is only one node
foreach (count($body) ? $body : new Node([$body]) as $k => $n) {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

what if there is a single parsed node with children ? count($body) would be truthy so it would iterate over the children of that node instead of checking the body

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

maybe we need to check for Node::class === \get_class($node) instead, as done in the old filterBodyNodes

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

maybe we should introduce a StatementListNode (name to be bikeshed) to represent such list of other nodes to allow identifying this case more easily and deprecate instantiating the Node class itself (making it abstract in Twig 4)

throw new SyntaxError(sprintf('The "%s" tag must always be at the root of the body of a template.', $node->getNodeTag()), $node->getTemplateLine(), $node->getSourceContext());
}

if ($this->currentTagNode && $node instanceof BlockReferenceNode) {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

shouldn't this check $this->currentBlockNodeLevel ? Once we are inside a block, we can have block definitions anywhere as those are not top-level blocks anymore.

}

if ($this->currentTagNode && $node instanceof BlockReferenceNode) {
if ($this->currentTagNode instanceof NodeCaptureInterface || count($this->blockNodes) > 1) {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

what is the goal of this count($this->blockNodes) condition ? why would we have a behavior that depends on the number of blocks defined in the template ?

if (str_contains((string) $node, \chr(0xEF).\chr(0xBB).\chr(0xBF))) {
$t = substr($node->getAttribute('data'), 3);
if ('' === $t || ctype_space($t)) {
// bypass empty nodes starting with a BOM
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is the filtering of the BOM gone ?

foreach ($node as $k => $n) {
if (null !== $n && null === $this->filterBodyNodes($n, $nested)) {
$node->removeNode($k);
foreach ($body as $k => $node) {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What happens when $body is not a Node but a child class because subparse did not wrap it ? AFAICT, this logic would fail to remove the node in such case.

--TEST--
conditional "block" tag with "extends" tag (nested)
--DEPRECATION--
Since twig/twig 3.14: Having a "block" tag under a "if" tag (line 7) is deprecated in layout.twig at line 8.
Copy link
Member

@stof stof Sep 10, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is a tag inside a non-top-level condition, so I don't think we need to deprecate this. Such tag is useful when you want to reuse part of the rendering between the if and else clause by defining the block in the if and referencing it in the else.
Having to move it to a top-level block (that does not override a parent block) makes the intent a lot less clear as it is easy to miss that this block is related to the rendering inside the condition (and so that it can be removed if the condition body does not need it anymore).

For the same reason, it can make sense to define a block inside the body of a loop, which would probably also trigger such a deprecation.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

To me, once entering a top-level block of a child template, we should not have any restriction on blocks anymore, just like there is no such restriction in a standalone template.

$this->currentBlockNodeLevel = 0;
}
if ($this->hasParent && $node->getNodeTag() && !$node instanceof BlockReferenceNode) {
$this->currentTagNode = null;
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks wrong to me, as it will reset the current tag node to null rather than resetting it to its state before entering that node (try putting a block inside a nested if in a template followed by a block after the inner if but inside the outer one, and your current code would not report the second block as being incorrect)

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

or even just some other nested node inside the if followed by a block after the if

@@ -188,6 +188,29 @@ Templates
``Template::loadTemplate()``); pass instances of ``Twig\TemplateWrapper``
instead.

* Having a "block" definition nested in another node that captures the output
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not answering on the "why", i have'nt tried this PR locally yet... but i feel the before/after examples do not represent the same usage/need.

For instance (please don't be too harsh on this code 🙏 ) this is a pattern i used from time to time, and i feel this is often how the set/block was used.

https://github.com/symfony/ux/blob/8a66e74d5e7070e5cbf60042cae535fae28d3e7a/ux.symfony.com/templates/components/Code/CodeLine.html.twig

I don't mind we cannot in 4.0 and will find a better way, but the "after" here would not help in this situation, right ?

fabpot added a commit that referenced this pull request Sep 27, 2024
…ode and Nodes (fabpot)

This PR was merged into the 3.x branch.

Discussion
----------

Deprecate instantiating Node directly, introduce EmptyNode and Nodes

Based on some comments from `@stof`:

See #4292 (comment)
See #4333 (comment)

First interesting usage here: 65ee72a

Commits
-------

8b27898 Deprecate using Node directly, introduce EmptyNode and Nodes
@stof
Copy link
Member

stof commented Oct 14, 2024

@fabpot what is the state of this PR ? Do you plan to fix it or to abandon it ?

@fabpot
Copy link
Contributor Author

fabpot commented Oct 14, 2024

@fabpot what is the state of this PR ? Do you plan to fix it or to abandon it ?

I plan to still work on this (soon).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Development

Successfully merging this pull request may close these issues.

Conditional {% use %} not working
3 participants