Repository navigation
Conversation
…ePeriod, and derived types in general
…ceholders The placeholder operations only compiled with icpc, which does not check discarded `if constexpr` branches strictly. With g++ and clang: - referencePeriodEncoding.h: include the numberOfAdditionalParameters deduction and set_or_throw; the deductions that do not exist yet (numberOfReferencePeriodTimeRanges, the additional-parameter scale factors and values) become TODOs. - quantileEncoding.h: declare Mars2GribConceptException before it is used, and drop the unreachable code after the throw. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Post-processed products carry the `postproc` block of PDT 4.89, 4.90, 4.133 and 4.135 (inputProcessIdentifier, inputOriginatingCentre, typeOfPostProcessing). The concept is active for type=pfc (ecPoint post-processed quantile forecasts). The encoding is a placeholder and is never applied. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…reference-period products Section 4 recipes: - 5/9 standard probabilities, 121/122 strike probabilities (no reference period: the large-ensemble templates have no such block), 112/131 probabilities of (standardised) anomalies. - 86/87 quantiles, 89/90 post-processed quantiles (pfc), 132-135 quantiles of anomalies, 106/107 anomalies of members/derived products. - Add the missing Section 4 registry entries (5, 9, 86, 87, 89, 90, 106, 112, 132-135); without them the template number was never set. Matchers: - probability: type=ep only; StandardisedAnomaly 133093-133098, Anomaly 131001-131010/131020-131025, StandardProbability 131060-131081, 131085, 131098-131100, StrikeProbability 131089-131091. - ensemble: ProbabilityLargeEnsemble only for strike probabilities. - referencePeriod: efi/efic/sot, 133093-133098, ep anomaly probabilities, anomaly (171xxx, 173xxx) and significance (234xxx) parameters. - quantile: has `quantile` and type != sot. - statistics/pointInTime: probabilities that ecCodes defines without a statistic (131068, 131069, 131073, 131074-131081, 131089-131091) are point in time, or Maximum over a time window (GRIB1 timeRangeIndicator 2); 133093-133098 windows are Average (ECC-2332); anomaly probabilities and anomaly/significance parameters are time means (tp anomalies are accumulations). - level: anomaly probabilities (131001-131010), mean wave period thresholds (131078-131081), anomaly/significance parameters on the level of their base parameter. Fixes in the WIP: Select<ProbabilityConcept, ...>, `range` in the referencePeriod matcher, Anomaly in ProbabilityList, 131020 range, S4_R106 in the recipe list. New test mars2grib-test-recipes-probability-quantile checks the Section 4 template selection without encoding. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
68 headers under backend/deductions and backend/checks use the dictionary traits but did not include them, so they only compiled when an earlier include had pulled in dictionary_access_traits.h. Include it where it is used. Each header now compiles on its own with g++. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The key was removed twice, so the second remove threw, and the success path fell through to mars2marsUnreachable() without returning. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MARS `quantile` is `<quantileValue>:<totalNumberOfQuantiles>` (e.g. `10:100`), the form of the ecCodes `mars.quantile` alias. It is built from the section-4 keys, so it does not depend on the ecCodes MARS definitions of the input stream/type. Needed for quantile products such as enfo pfc. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…antile` In GRIB1 (local definition 19) the SOT percentile is MARS `number` (90 or 10), with the outer percentile in `efiOrder` (99 or 1). In GRIB2 the two percentiles are reference-period parameters and ecCodes exposes them as MARS `quantile` (ECC-2000): SOT90 is `90-99:100`, SOT10 is `1-10:100`. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ducts For type efi/efic/sot, pass numberOfForecastsInEnsemble, efiOrder and the model climate of GRIB1 local definition 19 (numberOfReforecastYearsInModelClimate, sampleSizeOfModelClimate) to the misc dictionary, for the referencePeriod and derived concepts. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…Processing) Index products (EFI, SOT, ...) have two time ranges, typeOfStatisticalProcessing = [102, X] (ecCodes grib2/paramId.def). The ProductTimeSpec model already adds the outer IndexProcessing range (shape IFSFakeSingleLoopDoubleLoop); pass it the inner statistic X of the underlying field from a per-param table, and make the IndexProcessing variant applicable. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…roducts - typeOfGeneratingProcess: type=pfc is a post-processed forecast (13); efi/efic/sot are ensemble-derived products (4), as em/es. - derived ShiftOfTails: set numberOfForecastsInEnsemble (the SOT percentile is now MARS `quantile`, not `number`). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Set totalNumberOfQuantiles and quantileValue from the MARS `quantile` keyword (`<quantileValue>:<totalNumberOfQuantiles>`, validated by metkit::Quantile). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
For MARS type=pfc (ecPoint post-processed quantile forecasts) set the post-processing keys of the operational GRIB2 products: typeOfPostProcessing=206 (weather-type-subgrid-calibration), inputProcessIdentifier=16, inputOriginatingCentre=98. Other types throw. Test 012_enfo_pfc_quantile: pfc accumulation encodes as PDT 4.90 with the quantile and post-processing keys. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
numberOfDaysInClimateSamplingWindow (GRIB1 local definition 19) is the second reference-period time range of the model climate. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
StageAllocate:
- numberOfAdditionalParametersForReferencePeriod: 2 for SOT, else from
the parameter dictionary, else 0 (the deduction no longer throws).
- SOT: the two percentiles from MARS `quantile` (`90-99:100` -> {90, 99},
`1-10:100` -> {10, 1}, ECC-2000), checked against efiOrder. Set before
the local section, because ecCodes derives `mars.quantile` from them as
soon as `marsStream` is set.
- Model climate time ranges from the parameter dictionary:
numberOfReforecastYearsInModelClimate (years) and
numberOfDaysInClimateSamplingWindow (days).
StagePreset:
- efi/efic/sot: typeOfReferenceDataset=2 (reforecast),
typeOfRelationToReferenceDataset 20 (EFI) / 21 (SOT); for the other
products the relation follows from the paramId.
- sampleSizeOfReferencePeriod from sampleSizeOfModelClimate.
The start of the reference period is not encoded yet (TODO): it is not
part of the input metadata.
Test 013_enfo_index_products: efi, efic, SOT90 and SOT10 (PDT 4.107).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
probabilityType and the probability limits are defined per paramId in ecCodes and set with the `param` concept, so the probability concept has nothing else to encode. Drop the "not implemented" warnings that were printed for every strike, standardised-anomaly and anomaly field. Test 014_enfo_probabilities: tp probability over 24h (PDT 4.9), 10m wind speed probability instant (4.5) and over a time window (4.9, maximum), strike probability over 48h (4.122), probability of a standardised anomaly (4.131). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
typeOfGeneratingProcess had no mapping for type=ep, so the value of the sample (0, analysis) was kept. Use 5 (probability forecast). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## develop #295 +/- ##
===========================================
+ Coverage 50.96% 53.65% +2.69%
===========================================
Files 481 497 +16
Lines 22812 23391 +579
Branches 1569 1587 +18
===========================================
+ Hits 11625 12551 +926
+ Misses 11187 10840 -347 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
dsarmany
marked this pull request as draft
October 2, 2026 22:16
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
3 participants
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.
Description
Contributor Declaration
By opening this pull request, I affirm the following:
📋 Metkit Documentation 📋
https://sites.ecmwf.int/docs/metkit/pull-requests/PR-295