# Sources and Originality

MoonMissingness is an original MoonBit implementation for the August MoonBit
hackathon.

No third-party source code has been copied into this repository. The current
algorithms are standard missing-data techniques implemented directly in
MoonBit. The implementation is organized into the following original modules:

- `diagnose`: shape validation, missing-rate summaries, row/column statistics,
  joint-pattern reports, pairwise/conditional matrices, simulation, sampling,
  and engineering-oriented MCAR/MAR screening.
- `impute`: simple, KNN, sequence, regression, iterative/MICE-style, robust,
  ensemble, uncertainty, calibration, pipeline, and validation helpers.
- `csv`: CSV tokenization, quoted-field handling, numeric conversion, missing
  token normalization, schema profiling, filtering, and row transformations.
- `cmd/bench`: deterministic synthetic fixtures and benchmark checksums; the
  fixture is generated by the repository and is not copied from an external
  dataset.

The test files contain boundary fixtures written for this repository. They
cover empty tables, ragged rows, all-missing columns, duplicate rows, invalid
parameters, quoted CSV fields, deterministic masks, and metric edge cases.

The public interface files under package build directories are generated by
MoonBit's `moon info` command and are not hand-written source.

The project is released under Apache-2.0. Test fixtures are small synthetic
tables written specifically for this repository.

Related MoonBit ecosystem checks were performed on mooncakes.io using keywords
such as `missingness`, `impute`, and `missing data`. No mature package with the
same primary scope was found during this revision. Adjacent packages may cover
DataFrame storage, decision trees, or other analytics tasks; MoonMissingness is
scoped as a missing-data diagnosis and imputation layer.