+ 
+ 
+ 
+ 
+ + Turn merged pull requests and repository evolution into grounded, + executable tasks on healthy modern revisions. +
+ ++ Turn merged pull requests and repository evolution into grounded, + executable tasks on healthy modern revisions. +
+ +
- 
- 
- Turning Any Codebase into Testable Sandbox Environment + Change2Task workflow from the paper: historical evidence → modern base → L1/L2/L3 construction → H/C/R validation.
-RepoLaunch is the **first** agentic SWE tool for repository build and test management across programming languages and operating systems. Given a code repository, RepoLaunch can: -- Install all dependencies and build the repository, delivered as a docker image, with docker image layer info collected for Dockerfile reconstruction; -- Organize the command to rebuild the repository inside a container after repo modifications; -- Organize command to test the repository, write a parser to parse test output into structured testcase-status mapping, and optionally find per-testcase running command. +--- + +## Why Change2Task? + +An executable coding-agent task is more than an issue description. It needs a +specific repository state, a realistic objective, development tools, protected +behavior, and a verifier that distinguishes success from plausible-looking +failure. + +Change2Task treats repository history as reusable task evidence. It transfers a +historical maintenance change onto a runnable modern revision, constructs a +challenge state, and verifies the complete lifecycle: + +```text +Healthy H ──task patch──▶ Challenge C ──restoration──▶ Restored H′ + target ✓ target ✗ target ✓ + regression ✓ regression ✓ regression ✓ +``` + +RepoLaunch established the first chapter of environment engineering: making +repositories buildable and testable. Change2Task is the sequel: obtaining more +verified coding-agent tasks from every maintained environment. + +## What is included? + +|
+ Construction pipeline + L1 Patch Reversal, L2 Code Mapping, and bounded L3 Agent Reconstruction + behind one shared acceptance contract. + |
+
+ Five task families + Bug Fix, Feature Addition, Test Generation, API Migration, and Security + Repair. + |
+
+ Public task corpus + 900 provenance-linked task pairs from 252 public GitHub repositories and + 12 benchmark sources. + |
+
|
+ Strict validation + Qualification, scope, fidelity, repeated H/C/R lifecycle, explicit + infrastructure failures, and append-only evidence. + |
+
+ Agent-ready interface + A typed Python API, the change2task CLI, project Skill,
+ AGENTS.md, and one-command Skill installation.
+ |
+
+ Release hygiene + Portable paths, privacy review, source provenance, pinned-base license + inventory, strict schema, and complete checksums. + |
+
{escape(repr(self))}\"\n return formatting_html.array_repr(self)\n \n- def __format__(self: Any, format_spec: str = \"\") -> str:\n- if format_spec != \"\":\n- if self.shape == ():\n- # Scalar values might be ok use format_spec with instead of repr:\n- return self.data.__format__(format_spec)\n- else:\n- # TODO: If it's an array the formatting.array_repr(self) should\n- # take format_spec as an input. If we'd only use self.data we\n- # lose all the information about coords for example which is\n- # important information:\n- raise NotImplementedError(\n- \"Using format_spec is only supported\"\n- f\" when shape is (). Got shape = {self.shape}.\"\n- )\n- else:\n- return self.__repr__()\n+ def __format__(self: Any, format_spec: str) -> str:\n+ # we use numpy: scalars will print fine and arrays will raise\n+ return self.values.__format__(format_spec)\n \n def _iter(self: Any) -> Iterator[Any]:\n for n in range(len(self)):\n", "changed_paths": ["xarray/core/common.py"], "contract": {"allowed_output": "implementation", "objective": "Repair the described faulty behavior."}, "instance_id": "rq2_princeton_nlp_swe_bench_028__prinjector__pydata__xarray-6823", "regression_tests": ["xarray/tests/test_formatting.py::TestFormatting::test_get_indexer_at_least_n_items", "xarray/tests/test_formatting.py::TestFormatting::test_first_n_items", "xarray/tests/test_formatting.py::TestFormatting::test_last_n_items", "xarray/tests/test_formatting.py::TestFormatting::test_last_item", "xarray/tests/test_formatting.py::TestFormatting::test_format_item", "xarray/tests/test_formatting.py::TestFormatting::test_format_items", "xarray/tests/test_formatting.py::TestFormatting::test_format_array_flat", "xarray/tests/test_formatting.py::TestFormatting::test_pretty_print", "xarray/tests/test_formatting.py::TestFormatting::test_maybe_truncate", "xarray/tests/test_formatting.py::TestFormatting::test_format_timestamp_invalid_pandas_format", "xarray/tests/test_formatting.py::TestFormatting::test_format_timestamp_out_of_bounds", "xarray/tests/test_formatting.py::TestFormatting::test_attribute_repr", "xarray/tests/test_formatting.py::TestFormatting::test_diff_array_repr", "xarray/tests/test_formatting.py::TestFormatting::test_diff_attrs_repr_with_array", "xarray/tests/test_formatting.py::TestFormatting::test_diff_dataset_repr", "xarray/tests/test_formatting.py::TestFormatting::test_array_repr_variable", "xarray/tests/test_formatting.py::test_inline_variable_array_repr_custom_repr", "xarray/tests/test_formatting.py::test_set_numpy_options", "xarray/tests/test_formatting.py::test_large_array_repr_length", "xarray/tests/test_formatting.py::test__mapping_repr[50-40-30]", "xarray/tests/test_formatting.py::test__mapping_repr[35-40-30]", "xarray/tests/test_formatting.py::test__mapping_repr[11-40-30]", "xarray/tests/test_formatting.py::test__mapping_repr[1-40-30]", "xarray/tests/test_formatting.py::test__element_formatter"], "restoration_apply_mode": "reverse", "restoration_patch": "diff --git b/xarray/core/common.py a/xarray/core/common.py\nindex d1aac063..4c506b05 100644\n--- a/xarray/core/common.py\n+++ b/xarray/core/common.py\n@@ -188,22 +188,9 @@ class AbstractArray:\n return f\"{escape(repr(self))}\"\n return formatting_html.array_repr(self)\n \n- def __format__(self: Any, format_spec: str = \"\") -> str:\n- if format_spec != \"\":\n- if self.shape == ():\n- # Scalar values might be ok use format_spec with instead of repr:\n- return self.data.__format__(format_spec)\n- else:\n- # TODO: If it's an array the formatting.array_repr(self) should\n- # take format_spec as an input. If we'd only use self.data we\n- # lose all the information about coords for example which is\n- # important information:\n- raise NotImplementedError(\n- \"Using format_spec is only supported\"\n- f\" when shape is (). Got shape = {self.shape}.\"\n- )\n- else:\n- return self.__repr__()\n+ def __format__(self: Any, format_spec: str) -> str:\n+ # we use numpy: scalars will print fine and arrays will raise\n+ return self.values.__format__(format_spec)\n \n def _iter(self: Any) -> Iterator[Any]:\n for n in range(len(self)):\n", "target_tests": ["xarray/tests/test_formatting.py::TestFormatting::test_array_repr"], "task_statement": "RuntimeError when formatting sparse-backed DataArray in f-string\n### What happened?\n\nOn upgrading from xarray 2022.3.0 to 2022.6.0, f-string formatting of sparse-backed DataArray raises an exception.\n\n### What did you expect to happen?\n\n- Code does not error, or\n- A breaking change is listed in the [“Breaking changes”](https://docs.xarray.dev/en/stable/whats-new.html#breaking-changes) section of the docs.\n\n### Minimal Complete Verifiable Example\n\n```Python\nimport pandas as pd\nimport xarray as xr\n\ns = pd.Series(\n range(4),\n index=pd.MultiIndex.from_product([list(\"ab\"), list(\"cd\")]),\n)\n\nda = xr.DataArray.from_series(s, sparse=True)\n\nprint(f\"{da}\")\n```\n\n\n### MVCE confirmation\n\n- [X] Minimal example — the example is as focused as reasonably possible to demonstrate the underlying issue in xarray.\n- [X] Complete example — the example is self-contained, including all data and the text of any traceback.\n- [X] Verifiable example — the example copy & pastes into an IPython prompt or [Binder notebook](https://mybinder.org/v2/gh/pydata/xarray/main?urlpath=lab/tree/doc/examples/blank_template.ipynb), returning the result.\n- [X] New issue — a search of GitHub Issues suggests this is not a duplicate.\n\n### Relevant log output\n\n```Python\n# xarray 2022.3.0:\n\n{escape(repr(self))}\"\n return formatting_html.array_repr(self)\n \n- def __format__(self: Any, format_spec: str) -> str:\n- # we use numpy: scalars will print fine and arrays will raise\n- return self.values.__format__(format_spec)\n+ def __format__(self: Any, format_spec: str = \"\") -> str:\n+ if format_spec != \"\":\n+ if self.shape == ():\n+ # Scalar values might be ok use format_spec with instead of repr:\n+ return self.data.__format__(format_spec)\n+ else:\n+ # TODO: If it's an array the formatting.array_repr(self) should\n+ # take format_spec as an input. If we'd only use self.data we\n+ # lose all the information about coords for example which is\n+ # important information:\n+ raise NotImplementedError(\n+ \"Using format_spec is only supported\"\n+ f\" when shape is (). Got shape = {self.shape}.\"\n+ )\n+ else:\n+ return self.__repr__()\n \n def _iter(self: Any) -> Iterator[Any]:\n for n in range(len(self)):\n", "target_tests": ["xarray/tests/test_formatting.py::TestFormatting::test_array_repr"], "task_statement": "RuntimeError when formatting sparse-backed DataArray in f-string\n### What happened?\n\nOn upgrading from xarray 2022.3.0 to 2022.6.0, f-string formatting of sparse-backed DataArray raises an exception.\n\n### What did you expect to happen?\n\n- Code does not error, or\n- A breaking change is listed in the [“Breaking changes”](https://docs.xarray.dev/en/stable/whats-new.html#breaking-changes) section of the docs.\n\n### Minimal Complete Verifiable Example\n\n```Python\nimport pandas as pd\nimport xarray as xr\n\ns = pd.Series(\n range(4),\n index=pd.MultiIndex.from_product([list(\"ab\"), list(\"cd\")]),\n)\n\nda = xr.DataArray.from_series(s, sparse=True)\n\nprint(f\"{da}\")\n```\n\n\n### MVCE confirmation\n\n- [X] Minimal example — the example is as focused as reasonably possible to demonstrate the underlying issue in xarray.\n- [X] Complete example — the example is self-contained, including all data and the text of any traceback.\n- [X] Verifiable example — the example copy & pastes into an IPython prompt or [Binder notebook](https://mybinder.org/v2/gh/pydata/xarray/main?urlpath=lab/tree/doc/examples/blank_template.ipynb), returning the result.\n- [X] New issue — a search of GitHub Issues suggests this is not a duplicate.\n\n### Relevant log output\n\n```Python\n# xarray 2022.3.0:\n\n