--- tags: - code - go - code-style-analysis - multi-label-classification license: mit language: - go source_datasets: - bigcode/the-stack-v2 task_categories: - text-classification task_ids: - multi-label-classification dataset_info: features: - name: code dtype: string description: A snippet of Go source code. - name: labels dtype: sequence: class_label: names: - assignOp - builtinShadow - captLocal - commentFormatting - elseif - ifElseChain - paramTypeCombine - singleCaseSwitch description: > One or more style-rule violations detected by the go‑critic linter's "style" checker group. splits: - name: train num_examples: 1536 - name: validation num_examples: 222 - name: test num_examples: 448 dataset_size: 2206 configs: - config_name: default data_files: - split: train path: data/train-* - split: validation path: data/validation-* - split: test path: data/test-* --- # go-critic-style A **multi‑label** dataset of Go code snippets annotated with style violations from the [go‑critic linter's "style" group](https://go-critic.com/overview.html#checkers-from-the-style-group). Curated from the [bigcode/the‑stack‑v2‑dedup](https://huggingface.co/datasets/bigcode/the-stack-v2-dedup) "Go" split, filtered to examples of manageable length. ## Label Set List of style violations covered by this dataset: | ID | Label | Description | |--:|----------------------|---------------------------------------------------------------------| | 0 | `assignOp` | Could use `+=`, `-=`, `*=`, etc. | | 1 | `builtinShadow` | Shadows a predeclared identifier. | | 2 | `captLocal` | Local variable name begins with an uppercase letter. | | 3 | `commentFormatting` | Comment is non‑idiomatic or badly formatted. | | 4 | `elseif` | Nested `if` statement that can be replaced with `else-if`. | | 5 | `ifElseChain` | Repeated `if-else` statements can be replaced with `switch`. | | 6 | `paramTypeCombine` | Function parameter types that can be combined (e.g. `x, y int`). | | 7 | `singleCaseSwitch` | Statement `switch` that could be better written as `if`. | ## Splits The dataset is partitioned into training, validation, and test subsets in a 70/10/20 ratio: | Split | # Examples | Approx. % | |---------------:|-----------:|----------:| | **train** | 1536 | 70% | | **validation** | 222 | 10% | | **test** | 448 | 20% |