[edkerk]

Where GECKO and geckopy gave different results for the same input, both now behave alike. Invalid input raises, data that cannot apply to an item is skipped with a warning, and an explicitly empty list selects nothing. Each change has a unit test (tc0300 to tc0700).

Function	Behaviour
loadKcatSource	a field that opens a quote and never closes it raises loadKcatSource:quote; a row of empty quoted fields is skipped
collectKcats, mergeKcats	a file named by both database_top and database_bottom is one candidate list; a leading okp_ of a source is dropped for matching
readDLKcatOutput	inf and malformed kcats are dropped; a row with a substrate the model lacks is dropped with a warning, and the option strict raises instead; a file without any kcat raises readDLKcatOutput:noKcats
writeDLKcatInput	an empty selection writes an empty file; the selection is a logical vector
kcatAssignmentConfig	an empty preset, a criteria other than max, min, median or mean, and a sources that is not a cell raise kcatAssignment:config
kcatsFromECTree	an explicitly empty ecRxns selects nothing; a repeated reaction gives one row
getStandardKcat	Pseudoreaction and SLIME rxn match in any case; a second call on a light model works
assignKcatValues	ec.source is the lower-case label with (wc=…, origin=…) when known, as geckopy writes it
resolveTunableContext	a source such as combined (sd=0.10, n=2) belongs to the group of its name before the bracket and gets that group's sigma0
ecStatsFromBrenda, loadBrendaMeasurements	per-EC median and maximum from loadBRENDAdata, EC keys keep their case; measurements.tsv.xz is read in preference to measurements.tsv
getECfromDatabase	the action display warns and takes the lowest-index protein; an empty logical selection searches nothing; any other action than display or ignore raises getECfromDatabase:unsupportedAction
ecFSEOF	targets are ranked by slope times correlation rounded to 9 significant digits, so scores equal up to solver noise keep model order
getECfromGEM	a model without model.eccodes warns (getECfromGEM:noEccodes) and every EC number is empty
findMetSmiles	the last row of a duplicated name wins; an empty metabolite name is skipped with a warning
fillCarbonNum	a model without metFormulas warns and gives every exchange the carbon number 1
addNewRxnsToEC, applyComplexData	a reaction whose enzymes already exist is added; a stoichiometry of 0 counts as 1
calculateFfactor, relaxProteomicsGreedy	{} means no enzymes, [] means every enzyme
constrainEnzConcs, fillEnzConcs	a negative concentration or abundance raises
constrainFluxData, loadFluxData	a missing exchange reaction, an invalid maxMinGrowth, an unknown condition, a missing file, a file without exchange columns and a ragged row raise
getEnzymeUsage, reportEnzymeUsage	an enzyme without a usage reaction is left out with a warning; zero=false works; a negative topAbsUsage raises
truncateValues	Inf is returned unchanged
applyKcatConstraints	on a light model the update list takes cobra ids and ec ids; all-zero kcats clear the pool coefficients
saveEcModel	the default metaData description is written when the model has none, and an existing one is kept

The format change for ec reactions without enzymes (no enzymes entry) is in RAVEN (writeYAMLmodel), which saveEcModel uses; this branch has its test, tc0700. Parity: CompMetEng/raven-gecko-parity ledger and scenarios.

[github-actions]

Unit test results

71 tests   71 ✅  11s ⏱️
 1 suites   0 💤
 1 files     0 ❌

Results for commit 87017dc.