lookml-parser/changelog.md

Parse LookML

# Changelog

⚡ indicates a breaking change

- v2
	- ⚡ removed the `_name` property on objects.
	- Added a `_type` property on all objects
	- As a replacement for `_name`, added multiple `_<type>: <name>` properties (e.g. `{ ... , "_type": "join", "_join": "users", "_explore": "orders"}` )
	- Added a `_n` property which gives the lexical position of the object in the original text
- v2.2
	- The return now contains model objects which have recursively included other referenced files' contents
	- In interactive mode, the output is slightly more compact
- v3.0
	- ⚡ yesno fields are now mapped to true/false
	- Added conditional comment feature. The CLI flag is --conditional-comment=FOO or -c FOO
- v4.0
	- ⚡ The grammar now recognizes attributes whose type starts with 'expr' as doubleSemiBlock-valued. If you were previously using such a naming convention in conditional-comment based LookML, you will need to use a new naming convention, or use doubleSemiBlock syntax. E.g. `expression_custom_filter: ${field} ;;`
	- ⚡ The grammar no longer ignores leading whitespace in a doubleSemiBlock.
- v5.0
	- ⚡⚡⚡ For simplicity and reduced output size, the output no longer includes the array version of collections of named objects. Only the object version, keyed by name, is returned.
- v5.1
	- Added the ability to control the output format of the files collection
- v6.0
	- ⚡ The _file_rel property has changed slightly and no longer includes the file's extension
	- Added a 'none' file output mode for use cases that only care about assembled models
	- Added support for subdirectories and corresponding include statements (Does not yet support remote imports)
- v6.1
	- Added a `cwd` parameter to conveniently run the parser relative to a specified working directory
- v6.2
	- More consistent handling of manifest file across `file-output` modes
- v6.3
	- Support for new LookML syntax: refinements, and maplists (e.g. `filters:[field:">0"]`)
- v6.4 (should have been major change)
	- ⚡ "metadata" naming convention changed. Metadata properties now start with `$` rather than `_`
	- ⚡ Nodes in the JSON representation no longer contain their own name in a variadic property name, it is now always in `$name`
	- The parser can now capture text metadata into the `$strings` property, supporting use cases such as comment analysis, whitespace analysis and round-tripping/generation. Strings metadata is removed by default for the CLI, but can be exposed with the `transform` parameter.
	- [beta] Ability to generate LookML strings from JSON data containing the captured string metadata. Only exposed as a module export from the `generate` subdirectory
- v6.5
	- More permissive model file naming. Any character now accepted in model names.
	- Added guards against prototype pollution
- v6.6
	- NPM ignore test-projects, which was growing in size and is normally not useful to consumers of the package
	- Added the ability to resolve extensions/refinements within models, by passing `transformations:{applyRefinementsExtensions:true}` to parseFiles
	- Update tests which were broken by a change in a dev dependency (deep-object-diff)
- v6.8
	- Add support for `include: "//..."`, a.k.a. imported projects
- v6.8.2
	- Fix handling of '..' in include patterns
	- Fix error when including one model file into another model file
- v6.9
	- Support reading YAML-based LookML Dashboard files 
- v6.9.1
	- YAML-based LookML Dashboard files correctly includable in model objects
	- YAML parsing documentation
- v6.10
	- Compact conditional comments (without linebreak on first line)
- v6.11
	- Whitespace preservation in front of conditional comments
- v6.11.1
	- Project-root imports (single leading slash) from context of imported projects now resolve correctly
- v7.0.0
	- Adds the ability to output position data for LookML objects, using the addPositions (p) transformation.
	- For Node.js module users, the getPositions function is also exposed to be used with individual string parsing results
	- ⚡ Removes the additional "model" object inserted into the top-level of model files. The information conveyed by this object was of minimal use, is better served by the model property at the project level (as exposed by the assembe models transformation), and introduced confusing and inconsistent behavior for use cases expecting direct representations of file contents.
	- ⚡ Removes the ancestry metadata (e.g. $model, $view, $explore) added to child objects. This convenience information is easy for consumers to reproduce, and the implementation for this metadata interacted with the above change. In the future, it could be re-implemented as an opt-in transformation, if any demand for it exists.
	- ⚡ Switches the default file representation in parseFiles output from by-type to by-name
	- ⚡ parseFiles now defaults to models applying extensions/refinements and removing abstract declarations 
- v7.0.1
	- Fix `getPositions` export for Node.js module users