Conversation
- Implement splitAssignmentsByVariable() using existing parser AST - Support assign, assignVar, parseAssign tags - Handle if/elsif/else, unless, for, case, capture, raw blocks - Prune empty branches when pruneEmptyBlocks is true - Preserve control-flow structure per variable - Add unit tests and sample fixtures - Export from index.js Co-authored-by: Sanket Mishra <sanket0896@users.noreply.github.com>
|
Cursor Agent can help with this pull request. Just |
![baz-reviewer[bot]](https://avatars.githubusercontent.com/in/933528?s=60&v=4){kind=small}
| if (t.name === 'comment' && !options.includeComments) continue | ||
| if (t.name === 'raw') { | ||
| if (keepNonAssign) result.push(t) | ||
| continue | ||
| } | ||
|
|
||
| const filtered = filterControlFlow(t, targetVar, options) | ||
| if (filtered) { | ||
| if (pruneEmptyBlocks && isEmptyBlock(filtered)) continue | ||
| result.push(filtered) | ||
| } | ||
| } |
There was a problem hiding this comment.
renderTemplateToLiquid doesn't handle the comment tag, so comments preserved by filterTemplates when includeComments: true (line 122) fall through to the final return ''. As a result includeComments never renders comment tags and the output Liquid drops them. Can we add an if (name === 'comment') return (token && token.raw) || '' branch (similar to assigns)?
Finding type: Logical Bugs | Severity: 🟠 Medium
Want Baz to fix this for you? Activate Fixer
Other fix methods
Prompt for AI Agents:
In src/split-assignments.js around lines 122 to 344, the renderTemplateToLiquid function
fails to handle the 'comment' tag even though filterTemplates can keep comments when
includeComments is true; as a result comments are dropped from the output. Modify
renderTemplateToLiquid to add a branch for when name === 'comment' that returns (token
&& token.raw) || '' (similar to the existing assign-handling branch) so comment tags are
preserved in the rendered Liquid.
| function renderTemplateToLiquid (template) { | ||
| if (!template) return '' | ||
| if (template.type === 'html') return template.raw || template.value || '' | ||
| if (template.type === 'value') return (template.token && template.token.raw) || '' | ||
| if (template.type !== 'tag') return '' | ||
|
|
||
| const name = template.name | ||
| const token = template.token | ||
|
|
||
| if (['assign', 'assignVar', 'parseAssign'].indexOf(name) >= 0) { | ||
| return (token && token.raw) || '' | ||
| } |
There was a problem hiding this comment.
renderTemplateToLiquid only emits raw output for tag names ['assign','assignVar','parseAssign'], but filterTemplates can retain any tag listed in options.assignTags. A custom assignment tag (e.g. set) survives filtering but hits renderTemplateToLiquid's default case and becomes an empty string, so the per-variable slice silently loses the assignment; can we make rendering use options.assignTags (or a shared list) so filtered custom assignment tags are re-serialized instead of dropped?
Finding type: Logical Bugs | Severity: 🔴 High
Want Baz to fix this for you? Activate Fixer
Other fix methods
Prompt for AI Agents:
In src/split-assignments.js around lines 253 to 264, the renderTemplateToLiquid function
hardcodes ['assign','assignVar','parseAssign'] and so drops custom assignment tags that
passed filterTemplates. Make rendering aware of the configured assignTags by refactoring
renderTemplatesToLiquid and renderTemplateToLiquid to accept an options (or assignTags)
parameter and use options.assignTags instead of the hardcoded array. Propagate this new
parameter from the top-level splitAssignmentsByVariable where
renderTemplatesToLiquid(filtered) is called (change to renderTemplatesToLiquid(filtered,
opts)), and update all internal calls to pass the options down so custom assignment tags
are serialized rather than omitted.
| const impl = template.tagImpl | ||
| if (name === 'if' && impl && impl.branches && impl.branches.length > 0) { | ||
| let out = (token && token.raw) || '' | ||
| out += renderTemplatesToLiquid(impl.branches[0].templates || []) | ||
| for (let i = 1; i < impl.branches.length; i++) { | ||
| const b = impl.branches[i] | ||
| out += `{% elsif ${b.cond} %}` | ||
| out += renderTemplatesToLiquid(b.templates || []) | ||
| } | ||
| if ((impl.elseTemplates || []).length > 0) { | ||
| out += '{% else %}' |
There was a problem hiding this comment.
renderTemplateToLiquid hardcodes out += {% elsif ${b.cond} %}, so original {%- elsif … -%} or {% elsif … -%} whitespace-control modifiers are lost. This drops the trim directive and can change rendered HTML/newlines, violating keepNonAssign's intent to preserve original control-flow syntax. Can we store each branch's raw token when parsing and reuse it here instead of emitting a generic {% elsif … %}?
Finding type: Logical Bugs | Severity: 🔴 High
Want Baz to fix this for you? Activate Fixer
Other fix methods
Prompt for AI Agents:
In src/split-assignments.js around lines 266 to 276, the renderTemplateToLiquid function
emits a hardcoded `{% elsif ${b.cond} %}`, which loses any original whitespace-control
(e.g. {%- elsif … -%}) and breaks preservation of original control-flow formatting.
Modify the code so it uses the branch's original raw token when available: (a) in
filterControlFlow where newBranches are created (around lines ~147-166), copy the
original branch token (e.g. b.token or b.token.raw) onto each new branch object you
push; (b) in renderTemplateToLiquid (lines 266-276), replace the hardcoded `{% elsif ...
%}` emission with using b.token.raw (falling back to the current synthetic `{% elsif
${b.cond} %}` if b.token or b.token.raw is missing). This preserves any trim/whitespace
modifiers from the source.
2 participants
User description
Add
splitAssignmentsByVariablefunction to split Liquid templates by variable assignments, enabling granular processing of templates.Slack Thread
Generated description
Below is a concise technical summary of the changes proposed in this PR:
Enable splitting Liquid templates by assigned variable through the new splitAssignmentsByVariable helper, which parses via the factory’s Liquid engine and prunes AST branches so downstream processors can target each variable slice. Add regression coverage that exercises nested control flow, raw blocks, and ordering expectations to keep the per-variable slicing behavior stable.
Modified files (3)
Latest Contributors(0)
Modified files (2)
Latest Contributors(2)