# CSV Regression Fixtures — free preview 1.0.0

Three unchanged cases from the original 24-case full pack:

* `03-quoted-multiline`: a quoted LF does not start a new logical record.
* `07-blank-vs-missing`: empty strings, literal spaces and absent cells differ.
* `13-utf8-bom`: the declared UTF-8-sig decoder removes one leading BOM.

Python 3.10+ standard library only. From the extracted preview ZIP:

```sh
python3 validate.py
```

From the full source tree, use `python3 validate.py preview/manifest.json`.
Each manifest entry includes literal expected header/data values, SHA256,
delimiter, decoding policy and selected quality facts. `schema.json` describes
the format. Use these same inputs and expectations with your own parser.

The reference profile is Python `csv.reader(strict=True)`, doubled quote
escaping, no initial-space skipping and newline-neutral input. Missing/extra
counts compare data width with header width; empty present cells are exactly
empty strings. No null/number/Unicode normalization is performed. Parsing and
optional application limits are separate. This is a regression profile, not
a universal CSV standard, certification or performance benchmark.

New original synthetic data, code and documentation are MIT licensed; see
`LICENSE` and `RIGHTS.md`. No customer uploads, private records, third-party
datasets or runtime dependencies are bundled. The full pack adds 21 cases,
the generator, tests and reproducible release builder, with the same MIT
license. A suggested full-download price is an untested proposal, not income
or evidence of customers. No network or live API is required.