Skip to content

Latest commit

 

History

History
32 lines (18 loc) · 3.3 KB

codetf.md

File metadata and controls

32 lines (18 loc) · 3.3 KB
Raw

Code Transformation Format (CodeTF)

This describes Code Transformation Format (CodeTF).

This open format describes code changes or suggestions made by an automated tool. Much how the SARIF format published by Microsoft helped static analysis tools drive consistent interoperation, we believe an open format can benefit developers to begin contributing towards platforms that make and suggest changes, as well as reduce the need for vendor-specific viewers, parsers, etc. We hope to one day publish this specification through an open body like OASIS.

The specification JSON

The specification is expressed in terms of JSON Schema. The schema is currently not versioned. We are avoiding investment in ceremony, versioning, governance, etc., until we feel it has reached a more stable footing. Following SARIF stylistically as a long term goal makes sense, not only because it's a successful standard, but also because our results will be closely linked with SARIF, so we could have many users, consumers, and implementors in common.

Note that like SARIF, this format is not intended to be a replacement for a diagnostic log. It's not intended to have anything more than minimum diagnostics to help with reproducibility.

Structure

It may help to understand the major components of CodeTF from a high levels first before exploring or attemptin to implement the specification. The results and changeset fields can be seen as a series of patches against a project's directory. Each patch builds on any previous patches seen. Therefore, applying a patch from the middle of a changeset without the others may be invalid. Multiple locations can be changed in a single file within the scope of a single codemod and be represented by a single changeset array entry.

Codemod URIs

Codemods are uniquely identified by a URI, which is represented in CodeTF as the codemod property of the result object.

URIs are descriptive and must conform to the following schema: <detector>:<language>/<name>

Each component of the URI has a particular meaning:

  • <detector>: Describes the source of the analysis that drives the transformation. Codemods that remediate issues found by a specific analysis tool should be labeled with the detector corresponding to that tool name (e.g. "semgrep", "codeql", etc.). Implementers of custom codemods that perform their own internal detection should use a unique identifier for their detector. For example, Pixee's "find and fix" codemods use "pixee".
  • <language>: The language that is transformed by the codemod. This should be a short, unique identifier for the language. Valid languages include java, python, and javascript.
  • <name>: The name of the codemod. This should be a short, unique identifier for the transformation that is performed. Individual words in the name should be separated by hyphens. For example: remove-unused-imports.

Notes

Note that the changeset array can have multiple entries for the same given file.

Note that the lineNumber field should be deterministic, and this can't be traded for accuracy. The purpose of the line number is only to capture a location to be used in a --path-include or --path-exclude argument.