Missing indentation level after multiline tag

[pawamoy]

Input:

...
    {% filter heading(
        heading_level,
        role="data" if attribute.parent.kind.value == "module" else "attr",
        id=html_id,
        class="doc doc-heading",
        toc_label=('<code class="doc-symbol doc-symbol-toc doc-symbol-attribute"></code>&nbsp;'|safe if config.show_symbol_type_toc else '') + attribute.name,
    ) %}

      {% block heading scoped %}
        ...
      {% endblock heading %}

      {% block labels scoped %}
        ...
      {% endblock labels %}

    {% endfilter %}
...

Output:

...
    {% filter heading(
    heading_level,
    role="data" if attribute.parent.kind.value == "module" else "attr",
    id=html_id,
    class="doc doc-heading",
    toc_label=('<code class="doc-symbol doc-symbol-toc doc-symbol-attribute"></code>&nbsp;'|safe if config.show_symbol_type_toc else '') + attribute.name,
    ) %}

    {% block heading scoped %}
      ...
    {% endblock heading %}

    {% block labels scoped %}
      ...
    {% endblock labels %}

  {% endfilter %}
...

Expected (regardless of how the opening filter tag is indented):

...
    {% filter heading(
    heading_level,
    role="data" if attribute.parent.kind.value == "module" else "attr",
    id=html_id,
    class="doc doc-heading",
    toc_label=('<code class="doc-symbol doc-symbol-toc doc-symbol-attribute"></code>&nbsp;'|safe if config.show_symbol_type_toc else '') + attribute.name,
    ) %}

      {% block heading scoped %}
        ...
      {% endblock heading %}
  
      {% block labels scoped %}
        ...
      {% endblock labels %}
  
    {% endfilter %}
...

Even better (indent continuation lines twice, and ending ) %} only once):

...
    {% filter heading(
        heading_level,
        role="data" if attribute.parent.kind.value == "module" else "attr",
        id=html_id,
        class="doc doc-heading",
        toc_label=('<code class="doc-symbol doc-symbol-toc doc-symbol-attribute"></code>&nbsp;'|safe if config.show_symbol_type_toc else '') + attribute.name,
      ) %}

      {% block heading scoped %}
        ...
      {% endblock heading %}
  
      {% block labels scoped %}
        ...
      {% endblock labels %}
  
    {% endfilter %}
...

Reasoning: the nesting seems visually more correct, while keeping the continuation lines (heading_level, role=..., etc.) indented compared to opening and closing markers.

[pawamoy]

I'm able to at least fix detection of opening tags with this change: OPENING_TAG = r"{%[-+]? *[#/]?(\w+).*?(?:[-+]?%})?" (in DjTXT).

[pawamoy]

Here's a partial patch:

diff --git a/djhtml/modes.py b/djhtml/modes.py
index d017da3..69a5a34 100644
--- a/djhtml/modes.py
+++ b/djhtml/modes.py
@@ -56,6 +56,8 @@ class BaseMode:
         mode = self
         src = self.source
 
+        closing_marker = True
+
         while True:
             if src.find("\n") > mode.MAX_LINE_LENGTH:
                 raise MaxLineLengthExceeded
@@ -77,6 +79,14 @@ class BaseMode:
                 # Create a token from the head.
                 token, mode = mode.create_token(head, raw_token + tail, line)
                 line.append(token)
+                if isinstance(token, Token.Open):
+                    if not token.text.endswith("%}"):
+                        closing_marker = False
+                elif isinstance(token, Token.Text) and not closing_marker:
+                    if token.text.endswith("%}"):
+                        closing_marker = True
+                    else:
+                        token.relative += 1
 
             if raw_token == "\n":
                 self.lines.append(line)
@@ -209,7 +219,7 @@ class DjTXT(BaseMode):
         "video": (" as ", None),
         "placeholder": (" or ", None),
     }
-    OPENING_TAG = r"{%[-+]? *[#/]?(\w+).*?[-+]?%}"
+    OPENING_TAG = r"{%[-+]? *[#/]?(\w+).*?(?:[-+]?%})?"
 
     def create_token(self, raw_token, src, line):
         mode = self

Partial because it indents like this:

{% filter hello(
        param1,
        param2,
    ) %}

    {% block world %}
        ...
    {% endblock world %}

{% endfilter %}

{% with hello(
        param1,
    param2) %}

    {{ world }}

{% endwith %}

First case good (filter), second case not so good (with). Not sure how to fix it.

[pawamoy]

Reproducible with if and set too:

{%- if predicate1
    and predicate2
    and predicate3
  %}
  hello
{%- endif -%}

{%- set ns = namespace(
    has_pos_only=False,
    render_pos_only_separator=True,
    render_kw_only_separator=True,
    annotation="",
    equal="=",
  ) -%}

(set is not handled by my patch above, because it's probably not detected as an opening tag)

Changed the title accordingly.

[pawamoy]

Running djhtml twice with my patch formats the input a first time as I expect it, then formats it again as if the patch wasn't there: I'm not sure if this is an issue with my patch or with djhtml more generally (non-predictable results?).

[anentropic]

Some more examples:

        {% do item_kwargs.update({
        "checkbox_kwargs": {
        "name": _checkboxes_name,
        "value": item.value,
        "id": _checkboxes_name + "--id-" + loop.index|str,
        }
        }) %}

    {{ mdc_textfield(
    name + "--filter",
    label="Filter",
    attrs={
    "@input.debounce": "(e)=>$data.filterValue=e.target.value",
    "x-init": "filteredCheckboxListFilterInit"
    }
    ) }}

I would love if there was something like a 'python' mode inside the multiline tags, akin to current js/css modes, so that function/macro calls could be formatted nicely, ideally with dict/list literals indented too