This PR was merged into the 1.x branch.
Discussion
----------
fixed edge case in the method cache for Twig attributes
closes#2447
Commits
-------
413be416 fixed edge case in the method cache for Twig attributes
* 1.x:
Wrong stdClass class name generates warning
added more is_object() safegaruds
Autoescape test addition; collision fix in Array Loader.
test if is_object before calling method_exists, fixes#2431
bumped version to 1.33.1-DEV
prepared the 1.33.0 release
This PR was submitted for the master branch but it was merged into the 1.x branch instead (closes#2448).
Discussion
----------
Wrong stdClass class name generates warning
Commits
-------
38987c93 Wrong stdClass class name generates warning
This PR was merged into the 1.x branch.
Discussion
----------
Fix#2431
See title
Commits
-------
a5c2da6d test if is_object before calling method_exists, fixes#2431
This PR was squashed before being merged into the 1.x branch (closes#2434).
Discussion
----------
Autoescape test addition; collision fix in Array Loader.
This PR makes two changes. It was created while testing a third issue that was discovered to be fixed in the latest version.
## Added JS to Autoescape: name test
This adds "index.js.html" to the autoescape:name test, proving the functionality of the JS escape system and the name guessing strategy.
## Fixed collision in Array Loader
While trying to run the above test, it was discovered that the Array Loader uses the content of the template file as the "key" for determining which compiled template to load.
As the above test uses identical content for both HTML and JS templates this led to the compiled HTML template always being loaded.
This change prepends the content with the filename, which vaguely follows the same mechanism as the "options" (which are appended after `getCacheKey` is called).
**This change might be a backwards-incompatible change** depending on your interpretation of how `Loader/Array` is intended to be used.
The class itself states:
> This loader should only be used for unit testing.
The assumption is it's OK for this to occur.
Test cases have been updated to take this key change into account, and an additional test has been added to confirm `Loader/Array` specifically has the capability to resolve two files with the same content to different keys.
Commits
-------
da82b41a Autoescape test addition; collision fix in Array Loader.
* 1.x:
updated CHANGELOG
Call clearstatcache() before calling is_dir().
bumped version to 1.33
fixed CHANGELOG
Make "length" filter and "empty" test consider __toString [Twig 1.x]
This PR was submitted for the master branch but it was merged into the 1.x branch instead (closes#2262).
Discussion
----------
Call clearstatcache() before calling is_dir().
The result of is_dir() is cached.
So call clearstatcache() before calling 2nd is_dir().
Commits
-------
96638c0d Call clearstatcache() before calling is_dir().
This PR was squashed before being merged into the 1.x branch (closes#2420).
Discussion
----------
Make "length" filter and "empty" test consider __toString [Twig 1.x]
Use case: When you have variables in your views that are actually objects but implement `__toString`, they feel like strings: For example, `{{ something }}` will make use of that to-string-conversion.
What does *not* work is
**a)** `{{ something | length }}`, because that will only have a meaningful return value for objects implementing `\Countable`. This interface, however, may have a totally different semantic/purpose for the object in question.
**b)** `{% if something is empty %}`, because
> empty checks if a variable is an empty string, an empty array, an empty hash, exactly false, or exactly null `[http://twig.sensiolabs.org/doc/2.x/tests/empty.html]`
... and obviously `something !== null` in this case.
For template designers, this may be surprising if they don't actually care about the object-or-string difference, they just "use" the variable.
This change tries to address this as it changes the behavior for such objects that have a `__toString` method and are *not* `\Countable`.
*Yes*, it's a BC break in edge cases:
For a), objects that implement a `__toString` but not `\Countable` would previously yield `1` for `{{ object | length }}`, and now would return the length of the string returned by `__toString`.
For b), testing (defined) variables that are objects implementing `__toString` and that return `''`, the test now is `false`.
Commits
-------
f5193e92 Make "length" filter and "empty" test consider __toString [Twig 1.x]
This PR was submitted for the 2.x branch but it was merged into the 1.x branch instead (closes#2427).
Discussion
----------
Fixed JS escaping for unicode characters with higher code points
Unicode characters with higher code points were being escaped incorrectly. When these characters are escaped, they should maintain their surrogate halves. Previously, Twig was dropping the first surrogate half.
https://mathiasbynens.be/notes/javascript-escapes#unicode-code-point
> The tetragram for centre symbol (𝌆) has code point U+1D306, so you could write it as \u{1D306}. For comparison, if you were to use simple Unicode escapes to represent this symbol, you’d have to write out the surrogate halves separately: '\uD834\uDF06'.
Commits
-------
7f913495 Fixed JS escaping for unicode characters with higher code points
This PR was submitted for the 2.x branch but it was merged into the 1.x branch instead (closes#2422).
Discussion
----------
Updated the vim-twig link
The other project is not maintained anymore and the author recommends this fork.
Commits
-------
955d5abd Updated the vim-twig link
This PR was merged into the 1.x branch.
Discussion
----------
PHPUnit 6 compatability
This should make builds on Traivs CI green again.
Commits
-------
0d791de6 PHPUnit 6 compatability
This PR was squashed before being merged into the 2.x branch (closes#2415).
Discussion
----------
Add error message when calling parent() in a block that doesn't exist
Related: #2309
When calling `parent()` in a block that isn't defined in the parent template, you get an error message that states `Block "%s" on template "%s" does not exist`, even if the block is wrapped in an `is defined` check, which is confusing and makes debugging hard.
This fix adds a better error message indicating that you should not call `parent()` when the parent block is not defined.
Before:
After:
Commits
-------
4587f676 Add error message when calling parent() in a block that doesn't exist
This PR was submitted for the 2.x branch but it was merged into the 1.x branch instead (closes#2421).
Discussion
----------
Fix codeblock in recipes documentation
Commits
-------
adba4c2a Fix codeblock
Fabien Potencier
Christophe SAUVEUR
dave-newson
Matthias Pigulla
Shigeki Morimoto
Mike Almond
Evaldo Junior
Christian Flothmann
Pierre du Plessis
