dsh-checkdigit

dsh-checkdigit

Check-digit mathematics toolbox for DeepSeek Harness (dsh): generate, validate, repair and detect check digits for 17 schemes — Luhn, Verhoeff, Damm, ISBN-10, ISBN-13, EAN-8, EAN-13, UPC-A, GTIN-14, ISSN, ISIN, CUSIP, SEDOL, IBAN, CAS Registry Number, ABA routing number and NPI — plus an ISBN-10 ⇄ ISBN-13 converter, a single-edit repair search ("what was this number most likely meant to be?") and a batch validator for whole columns of identifiers.

校验位数学工具箱：生成 / 验证 / 修复 / 识别 17 种校验位（Luhn、Verhoeff、Damm、ISBN、EAN、UPC、GTIN-14、ISSN、ISIN、CUSIP、SEDOL、IBAN、CAS 化学文摘号、ABA 美国银行路由号、NPI 美国医疗提供者号），另带 ISBN-10 与 ISBN-13 互转、「单字符修复」搜索工具、以及批量校验工具；零运行时依赖、纯本地算术。

Why

Language models routinely botch check-digit arithmetic: wrong card-number checksums, ISBNs that fail validation, IBANs with impossible check digits. Every algorithm here is implemented from its published specification with pure integer arithmetic — deterministic, offline, zero runtime dependencies — and cross-checked against published worked examples (Apple's US0378331005, the classic GB82 WEST 1234 5698 7654 32, the Luhn textbook example 79927398713, water's CAS number 7732-18-5, and more).

Install

dsh plugin --profile web add github:TYEclipse/dsh-checkdigit

Then add dsh-checkdigit to dsh.profile.bundles in your profile's package.json if it is not auto-activated, and restart dsh.

Tools

checkdigit_generate

Compute the check digit for a payload and return the complete identifier.

checkdigit_generate(scheme="luhn", payload="7992739871")
→ checkDigit: "3", complete: "79927398713"

Supported scheme values and their payloads:

scheme · payload · check digit · example complete luhn · digits (any length ≥ 1) · last digit · 79927398713 verhoeff · digits (any length ≥ 1) · last digit · 2363 damm · digits (any length ≥ 1) · last digit · 5724 isbn10 · 9 digits · 10th char (0-9 or X) · 0306406152 isbn13 · 12 digits · 13th digit · 9780306406157 ean8 · 7 digits · 8th digit · 96385074 ean13 · 12 digits · 13th digit · 4006381333931 upca · 11 digits · 12th digit · 036000291452 gtin14 · 13 digits · 14th digit · 09780306406157 issn · 7 digits · 8th char (0-9 or X) · 03785955 isin · 2 letters + 9 alphanumerics · 12th char (digit) · US0378331005 cusip · 8 chars from [A-Z0-9*@#] · 9th char (digit) · 037833100 sedol · 6 chars from [0-9] + consonants · 7th char (digit) · B15KXQ8 iban · 2-letter country + BBAN · chars 3–4 (two digits) · GB82WEST12345698765432 cas · 2–9 digits (hyphens ignored) · last digit · 7732-18-5 aba · 8 digits · 9th digit · 021000021 npi · 9 digits · 10th digit · 1234567893

Use it to build test card numbers, ISBNs, barcodes, securities identifiers, routing numbers or account numbers whose check digits actually verify.

checkdigit_validate

Verify the check digit of a full identifier. The scheme is auto-detected when omitted (CAS hyphenated → ISIN → 9-digit ABA → CUSIP → SEDOL (letter-leading) → IBAN → EAN-13 → UPC-A → ISSN (X-terminated) → EAN-8 → ISBN-10 → Luhn); pass scheme to force one. SEDOLs that are all digits and every NPI share their length with Luhn/ISBN-10 values, so those two have to be named.

checkdigit_validate(value="GB82WEST12345698765432")
→ valid: true, scheme: "iban", iban.country: "GB", iban.formatted: "GB82 WEST 1234 5698 7654 32"

checkdigit_validate(value="7732-18-5")
→ valid: true, scheme: "cas", checkDigit: "5"

checkdigit_validate(value="021000021")
→ valid: true, scheme: "aba", checkDigit: "1"

checkdigit_validate(value="79927398714")
→ valid: false, scheme: "luhn", checkDigit: "4", expected: "3",
  detail: "Luhn: check digit should be 3, but the value ends in 4"

Spaces, dashes and dots in the input are ignored (the spaced card form 4111 1111 1111 1111, the hyphenated 978-0-306-40615-7 and the printed 0378-5955 all validate). When a value is invalid, the result carries the expected check digit so callers can fix it programmatically.

checkdigit_repair

When a value fails, enumerate the single-character edits that make it valid — a corrected check character, a swapped adjacent pair, or one mistyped payload character. Every repair is confirmed by the scheme's own validator and ordered most-likely-first.

checkdigit_repair(value="79927398714")
→ scheme: "luhn", total: 12, checkDigit: "4", expected: "3"
  repairs[0]: { type: "check-digit", position: 11, from: "4", to: "3", value: "79927398713" }

checkdigit_repair(value="GB82WEST12345698765431")
→ scheme: "iban", total: 7
  repairs[0]: { type: "check-digit", position: 3, to: "12", value: "GB12WEST12345698765431" }   # keep the account
  repairs[6]: { type: "substitute", position: 22, from: "1", to: "2", value: "GB82WEST12345698765432" }  # keep the check digits

Both IBAN answers are legitimate — the caller can see whether the account digits or the check digits were more likely mistyped. total always reports the full count; the returned list is capped at 20 (truncated: true).

checkdigit_batch

Validate a whole column of identifiers in one call — 1 to 50 values, request order preserved. Each value is auto-detected unless a scheme is given for the whole list, and the report carries one row per value plus the counts and a per-scheme breakdown.

checkdigit_batch(values=["79927398713","79927398714","B15KXQ8","B15KXQ7","978-0-306-40615-7",
                         "GB82WEST12345698765432","not-an-id","1234567893"])
→ total: 8, validCount: 4, invalidCount: 3, unknownCount: 1
  byScheme: [luhn×2, sedol×2, ean13×1, iban×1, isbn10×1, unknown×1]
  results[3]: { index: 4, value: "B15KXQ7", scheme: "sedol", valid: false, checkDigit: "7", expected: "8" }

checkdigit_batch(values=["1234567893","1245319599","1234567890"], scheme="npi")
→ total: 3, validCount: 2, invalidCount: 1, unknownCount: 0

Use it for pasted lists and spreadsheet columns instead of one validation call per row. An empty list or more than 50 values is rejected with an error and no rows — the list is never truncated silently.

checkdigit_info

Describe the supported schemes: identifier lengths, payload formats, check-digit positions and worked examples. Call with a scheme for one scheme, or without arguments for all 17.

isbn_convert

Convert between ISBN-10 and ISBN-13 in both directions. The source check digit is verified first; the 978 Bookland prefix is prepended (10→13) or dropped (13→10) and the target check digit recomputed. Hyphen grouping from the input is preserved in formatted.

isbn_convert(isbn="0-306-40615-2")
→ valid: true, direction: "to13", converted: "9780306406157", formatted: "978-0-306-40615-7"

isbn_convert(isbn="978-0-306-40615-7")
→ valid: true, direction: "to10", converted: "0306406152", formatted: "0-306-40615-2"

isbn_convert(isbn="9791090636071")
→ valid: false, error: "ISBN-13 with the 979 prefix has no ISBN-10 equivalent; only the 978 Bookland prefix maps back to ISBN-10"

ISBN-10 check digits may be X (value 10): 0-8044-2957-X → 978-0-8044-2957-3.

Supported schemes

Scheme · Standard · Detail Luhn · ISO/IEC 7812 · Credit/debit cards, IMEI, loyalty cards Verhoeff · — · Dihedral-group checksum; catches all single errors and adjacent transpositions Damm · — · Quasigroup operation; full-value tally must be zero ISBN-10 · ISO 2108 · Mod-11; check digit may be X (= 10) ISBN-13 · ISO 2108 · EAN-13 arithmetic (Bookland) EAN-8 / EAN-13 · GS1 GTIN-8 / GTIN-13 · Retail barcodes GTIN-14 / SSCC-18 · GS1 GTIN-14 · Logistics case and pallet codes; same mod-10 weighting as EAN/UPC over a 13-digit payload ISSN · ISO 3297 · Serial publications; mod-11 check character may be X UPC-A · GS1 GTIN-12 · North American retail barcodes ISIN · ISO 6166 · Securities; letters expand A=10…Z=35, then Luhn CUSIP · ANSI X9.6 · North American securities; * @ # allowed; even positions doubled SEDOL · London Stock Exchange · UK/Ireland securities; 7 chars, weights 1,3,1,7,3,9, letters A=10…Z=35, vowels never issued IBAN · ISO 13616 · Mod-97 over letter-expanded rearranged string; BBAN length table for 75+ countries CAS Registry Number · CAS · Chemical substances; digit × position-from-right sum mod 10 ABA routing number · US Federal Reserve · Bank routing transit numbers; 3-7-1 weighted sum mod 10 NPI · CMS / US HHS · US healthcare provider identifiers; plain Luhn over the constant 80840 prefix, first digit 1 or 2

Development

pnpm install
pnpm build
pnpm test   # 119 tests
pnpm lint
python3 test/oracle/anchors.py   # independent anchors for every numeric assertion

Test expectations are not hand-computed: test/oracle/anchors.py re-implements each scheme from its published specification in Python and prints every anchor the suites assert, including the repair counts and positions. Run it (or pnpm test on a fresh checkout) to regenerate them.

License

MIT