Expand deck diagnostics: cross-keyword rules, value-type checks, and summary-vector coverage - #51
Merged
Conversation
Walk a local opm-tests clone (OPM_TESTS_DIR, default M:/gitroot/opm-tests), run computeDiagnostics on every deck/include, and write a triage report grouping suspected false positives by diagnostic type and keyword. Gated with describe.skip when the corpus is absent so normal test runs are unaffected. The generated corpus-report.md is gitignored.
User-defined quantity names begin with a data-type letter (A B C F G R S W) followed by 'U' (WUOPRL, FU_VAR1, WU_WBHP). They are user-defined and never in the keyword index, but appear legitimately as bare SUMMARY mnemonics and inside ACTIONX/UDQ bodies. Recognise them by shape in computeDiagnostics. Cuts unrecognised-keyword false positives on the opm-tests corpus from 525 to 287.
Decks routinely split sections across included files (e.g. RUNSPEC in the master deck, the remaining sections pulled in via INCLUDE 'BASE'). When the master file is analysed standalone we never see the section headers inside the include, so the tracked section is stale and SCHEDULE keywords written after the INCLUDE were wrongly flagged 'not valid in RUNSPEC'. Once an INCLUDE/IMPORT/GDFILE is seen, suppress the wrong-section check until the next explicit section header. Cuts wrong-section false positives on the opm-tests corpus from 54 to 8.
…ated OPM Flow allows a record's terminating '/' to appear on a later line than its values (MINPV / PINCH / SPECGRID / PLYSHLOG and many table keywords are written this way). The per-line check flagged every such record as missing its '/'. Track an 'open' record instead: a record with values but no '/' is held open; a later '/' (standalone or trailing) closes it, and only a record still open at the block boundary (next keyword, section header, or EOF) is flagged. This also matches how OPM Flow itself parses — tokens are read until a '/'. Cuts missing-record-terminator false positives on the opm-tests corpus from 157 to 0.
Region summary vectors can be qualified by a named FIP region set, written as <region-vector>_<NAME> (ROIP_ABC, RPR__ABC). The base is an R-prefixed vector in the index; the _<NAME> qualifier is user-defined so the full token is never in the index. Recognise this form so it is not flagged as an unknown keyword. Cuts unrecognised-keyword false positives on the opm-tests corpus from 287 to 212 (total 771 -> 256 since the harness was added).
VFPPROD / VFPINJ (and other variadic-record keywords like RSVD) are classified 'list' but their table-style final record is closed by its own per-record '/', with no separate standalone list terminator. They were wrongly flagged as missing the '/' to close the record list. Skip the list/array close-terminator check for variadic-record keywords. Cuts missing-list/array-terminator false positives on the opm-tests corpus from 27 to 12.
OPM Flow only recognises keywords in column 1, so an indented token that is not
a known keyword cannot be one — it is body content of the active block. A single
well name on its own indented line under a SUMMARY vector (' PROD2 /') was
matching the keyword pattern and, when the vector did not 'expect more records',
being mis-flagged as an unknown keyword. Treat such indented non-keyword tokens
as record body. Column-1 detection for indented *known* keywords is preserved.
Cuts unrecognised-keyword false positives on the opm-tests corpus from 212 to
196.
…to report Completion-level (trailing L: WOPRL = WOPR + L) and LGR-local (leading L: LWWIR = L + WWIR) summary vectors are systematic variants that opm-common does not enumerate. Recognise them when the stripped base is an indexed SUMMARY vector, which keeps genuine typos flagged. Also extend the corpus harness report with the full list of unique unrecognised keywords as an index-coverage worklist. Cuts unrecognised-keyword false positives on the opm-tests corpus from 196 to 164.
opm-common defines only the base region-number keywords KRNUM and IMBNUM, so real decks using the per-direction Cartesian forms (KRNUMX/Y/Z, IMBNUMX/Y/Z) were flagged as unknown keywords. Emit the directional variants from the base entry in the index build, and rebuild the bundled compact index. Cuts unrecognised-keyword false positives on the opm-tests corpus from 164 to 134.
opm-common groups summary vectors into PROBE families (WELL_PROBE, FIELD_PROBE, BLOCK_PROBE, …) that enumerate their concrete deck mnemonics in a deck_names list (1691 unique) plus a deck_name_regex for open-ended families (UDQ, tracer, water-cut). The build did not expand them, so real decks requesting vectors like WWIP/WGIP/BTEMP/WXMF were flagged as unknown keywords. - Build: expand each PROBE's deck_names into minimal recognised entries (name, sections, one-line summary; no size shape, so no arity/terminator checks), and emit the deck_name_regex families to data/summary_name_patterns.json. - Diagnostics: computeDiagnostics takes the anchored summary-vector patterns and recognises matching tokens; the extension and corpus harness load and pass them. Also delivers the deck_names aliases groundwork (plan item 4). Compact index grows to 3183 keywords (2647 KB, under the 3 MB budget). Cuts unrecognised-keyword false positives on the opm-tests corpus from 134 to 34 (total 163 -> 63).
Emit opm-common's requires/prohibits constraints into the keyword index and surface them as document-wide diagnostics: a 'X requires Y' warning when a required keyword is absent (suppressed when the deck pulls in INCLUDE files), and a 'X conflicts with Y' warning when a prohibited partner is present. Implements plan item 1 of issue #6.
Validate each record value against the matching item's opm-common value_type: flag a quoted string in an INT/DOUBLE slot and a decimal in an INT slot. Bare identifiers (UDA/UDQ references) and defaults are left alone, and the check is skipped for variadic-record table keywords. The enum (options[]) mismatch check from the plan is deliberately not implemented: the option sets are heuristically extracted from the manual and are frequently incomplete/abbreviated (WCONINJE lists WAT, not the valid WATER), which produced ~3000 false positives on the known-good opm-tests corpus. Options stay in the index for completions only. Also harden the item-1 requires check: suppress it for INCLUDE fragments (no section header) and for phase requirements under implicit-phase modes (CO2STORE/H2STORE), eliminating the remaining corpus false positives. Implements plan item 2 of issue #6.
# Conflicts: # vscode-extension/src/analysis.ts
Two source files carried embedded NUL (0x00) bytes where a space delimiter was intended inside template literals: the requires/prohibits pair key in analysis.ts and the type+keyword corpus key in corpus.test.ts. The NULs made git treat the files as binary. Replaced each NUL with a space; type-check and the full jest suite (281 tests) pass unchanged.
The README's feature list and Unreleased release notes covered only the boilerplate completion / quick-fix work; the value-type and requires/prohibits diagnostics and the keyword-recognition / terminator relaxations added in this branch were undocumented. Add a Diagnostics feature entry for each new check and an Unreleased section summarising the user-facing changes.
42 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
This batch of changes substantially expands the diagnostic engine and summary-vector index coverage for the OPM Flow deck editor.
Diagnostics
deck_namealiases and suppress opm-common family containers./for variadic-record list keywords (VFP, RSVD).INCLUDE.Summary-vector index
deck_names into summary-vector index entries.Tooling & docs
scripts/build_keyword_index.pywith unit tests.corpus.test.ts).