Reading JV-Data without losing source values

Codes.parse accepts a CodeTable and its exact ASCII representation. It returns an OfficialCode for both recognized and unknown values; Codes.isKnown, Codes.label and Codes.raw distinguish them. Incorrect width or non-ASCII input returns Error.

Codes.parse CodeTable.Grade "A" |> Result.map Codes.label
// Ok (Some "G1（平地競走）")
Codes.parse CodeTable.Margin " 12" |> Result.map Codes.raw
// Ok " 12"

Do not trim codes or convert them to integers. Grade " ", coat "01" and margin "1 " have significant spaces or leading zeros. All 19 official tables are represented. Unknown values retain their table and raw code so applications can handle later extensions.

RecordBytes.envelope reads the common eleven-byte header and copies the original bytes. RecordHeader retains the record ID, data-category code and raw creation date. Category meanings are record-specific; an all-zero creation date becomes None with its raw representation retained. Invalid dates produce RecordParseError, including the record, field, byte position and offending value.

RecordBytes.field, text, ascii, number, date and crlf are curried primitives with one-based byte positions. Text decoding preserves Japanese, full-width spaces and ASCII spaces. Numbers accept a sign only when requested, distinguish blank from zero, and apply an explicit decimal scale.

Text uses Windows CP932, including its documented duplicate mappings such as FA4A → Ⅰ and ED40 → 纊. These decode identically on portable and Windows targets. Undefined or truncated sequences still return a located RecordParseError; decoding does not guess another encoding or replace malformed text. RecordBytes.field, parse-error bytes and complete record Raw retain the original byte spelling even when two CP932 encodings represent the same Unicode character. See the Microsoft CP932 mapping.

RecordKinds.ofId supplies the official meaning, including O2 as quinella and O3 as wide. Unrecognized IDs, including unofficial H5, remain Unknown. Header recognition alone does not validate a record body or select a historical layout. See record migration for supported bodies and explicit historical selection. The previous parsers are isolated under Xanthos.Legacy.Records and are not the runtime default.