Repository navigation
Deck reader: accept fractional .freq ndec (documented), plus compat sweep edge cases #154
Description
Activity
This issue's
loom:buildingclaim was acquired by sweepsweep-issue-154-1791243629on hosthost-34209a6eat 2026-10-05T23:40:32.357611+00:00. This comment is a lease record (Issue #6179, Epic #6165) — its liveness signal is this comment's own forge-assignedupdated_at, never a timestamp embedded in this text. Seedefaults/docs/lease-record.mdfor the format contract this establishes, anddefaults/docs/lease-renewal.mdfor how the owning sweep keeps it fresh for the lifetime of its claim. Nothing reads this record yet (write-only, Phase 1) — a future phase will use it to decide reclamation of an abandoned claim.Curator enhancement — 2026-10-06
The existing problem, measured table, and acceptance criteria are retained. PR #161 is the implementation already in progress; coordinate follow-up there rather than starting a second implementation. PR #159 cross-references this issue from a repository documentation pass. Duplicate/cross-reference checking found those two open PRs and no duplicate issue. The other intake/approved deck-reader issues address different directives; no sibling consolidation is warranted. This remains part of #76, with no prerequisite implementation issue identified.
Verified corrections
- 2026-10-06, verified against
origin/mainate5d924abb67d149f1f3a066678078c36af2c3e7e(git showand symbol search):Builder::apply_freqinfasterhenry-cli/src/inp.rsreadsndecusingparse_count, and itsfrequency_sweeptakesusize.decade_frequenciesinfasterhenry-cli/src/main.rsseparately rejects fractional density for unequal endpoints. Both paths need the documented fractional-density rule; exact source line numbers differ from the local checkout.
Implementation Guidance
- Parse
.freqdensity as a finite floating-point value, validating negative density before single-frequency/DC early returns. Preserve integer-only validation for unrelated subdivision counts. - Keep native and compatibility policies explicit: compatibility changes endpoint inclusion and supplies the documented issue defaults with line-numbered warnings; native mode retains its errors. Preserve exact whole-decade points without clamping the last generated point.
- Distinguish a successfully parsed empty compatibility sweep from a missing
.freqdirective. Check the existing CLI behavior when there are no frequencies and when--freqsupplies an override. - Share the documented positive-density generation rule with
--freqto avoid maintaining divergent formulas. Keep compatibility-only deck warnings in the deck parser. - Implement from this issue's clean-room specification. The original measured FastHenry behavior is supplied evidence, not independently remeasured during this curation.
Affected Files
All paths below exist on the fetched
origin/maincommit named above.fasterhenry-cli/src/inp.rs— density parsing, native/compat sweep policy, diagnostics, empty-sweep handling, and parser unit tests.fasterhenry-cli/src/main.rs— duplicated command-line decade generation and empty/overridden sweep behavior.fasterhenry-cli/src/cli.rs—--freqhelp text if it currently requires integer density.fasterhenry-cli/tests/cli.rs— command-line fractional-density and sweep override regression coverage.fasterhenry-cli/README.md— document fractional command-line density.docs/fasthenry-compat.md— update.freqcompatibility rows as required by the existing acceptance criteria.
Test Plan
- Automated parser tests reproduce every supplied frequency-table row; use numerical tolerance for nonintegral exponents and exact equality for whole-decade points. Assert the last point is stepped rather than clamped.
- Test
1..9.995 ndec=1in both modes, equal endpoints, and the existing zero-frequency/DC behavior. - Test negative, zero, omitted, malformed, and nonfinite density values in both modes; verify errors/warnings carry the
.freqline number. Include equal endpoints so early returns cannot bypass invalid-density checks. - Test reversed bounds within and outside the compatibility slack, missing
.freq, empty compatibility sweeps, and--freqoverriding an empty sweep. - Exercise fractional
--freqin CLI tests and verify it follows the native positive-density rule. - Run the CLI test suite plus workspace formatting/lint checks. Review updated docs/help against the tested mode behavior.
- Manual verification: use a self-authored minimal deck with
fmin=1e3 fmax=1e7 ndec=0.5; confirm the output lists 1e3, 1e5, and 1e7 Hz and compatibility warnings appear only for the specified exceptional inputs.
Curation is specification preparation, not approval of PR #161 or verification of its reported test results.
- 2026-10-06, verified against
- added a commit that references this issue
on Oct 7, 2026
Problem
User's Guide §1.3.6: "ndec need not be an integer". Its own example
fmin=1e3 fmax=1e7 ndec=0.5solves 10³, 10⁵, 10⁷ Hz. We reject any non-integer ndec ('0.05' is not a valid ndec), which stops 4 of the 19 runnable example decks.Measured rule (FastHenry, self-authored decks): solve f_k = fmin·10^(k/ndec) for k = 0, 1, … while f_k ≤ 1.001·fmax. The last point is the stepped value and is not clamped to fmax.
Edge cases: an omitted ndec acts as 1. ndec=0 warns and becomes 0.01. A negative ndec loops forever. fmin > 1.001·fmax gives an empty sweep. fmin=0 is DC only (already supported). The 0.1% slack also differs from us for integer ndec:
1..9.995 ndec=1is 2 points in FastHenry and 1 in ours.Expected
--fasthenry-compatonly (undocumented quirks): the f ≤ 1.001·fmax inclusion rule; ndec=0 → 0.01 + warning; omitted ndec → 1 + warning; fmin > fmax → empty sweep + warning. These stay errors in native mode.Acceptance criteria
docs/fasthenry-compat.md.freqrows updated.Clean-room provenance
Specified from the public FastHenry User's Guide / nonuniform ground-plane manual and from black-box runs of a locally built FastHenry on self-authored decks (2026-10-05, research for #74). No FastHenry source or distributed example deck was consulted or quoted; implement from this spec and the public docs only (
CONTRIBUTING.md). Found bytools/fasthenry_compare.pyover the FastHenry example corpus (#74).Part of #76.