Command line¶

Eighteen scimappro modules ship an argparse front end, so an analysis step can be run from a shell script or a workflow manager without writing Python.

There are no console entry points — pyproject.toml declares no [project.scripts] — so each one is invoked with python -m and its module path, which is not always the same as the function name.

python -m scimappro.tl.spatial_distance \
    --data analysis.h5ad \
    --phenotype phenotype \
    --outputDir results

Available commands¶

Module path	Function
`scimappro.pp.mcmicro_to_scimap`	`pp.mcmicro_to_scimap`
`scimappro.pp.rescale`	`pp.rescale`
`scimappro.pp.log1p`	`pp.log1p`
`scimappro.pp.combat`	`pp.combat`
`scimappro.tl.phenotype`	`tl.phenotype`
`scimappro.tl.cluster`	`tl.cluster`
`scimappro.tl.umap`	`tl.umap`
`scimappro.tl.foldChange`	`tl.foldChange`
`scimappro.tl.spatial_distance`	`tl.spatialDistance`
`scimappro.tl.spatial_cooccurrence`	`tl.spatialCooccurrence`
`scimappro.tl.spatial_aggregate`	`tl.spatialAggregate`
`scimappro.tl.spatialProximityScore`	`tl.spatialProximityScore`
`scimappro.tl.spatialSimilarityLookup`	`tl.spatialSimilarityLookup`
`scimappro.tl.neighCount`	`tl.neighCount`
`scimappro.tl.neighExp`	`tl.neighExp`
`scimappro.tl.neighLDA`	`tl.neighLDA`
`scimappro.tl.neighNMF`	`tl.neighNMF`
`scimappro.pl.barplot`	`pl.barplot`

Every one of them prints its full option list with --help:

python -m scimappro.tl.phenotype --help

Shared conventions¶

Flag	Meaning
`--data PATH`	The `.h5ad` file or `.zarr` store to work on. `--adata` is accepted as an alias for scripts carried over from scimap.
`--sdataTable NAME`	Which table to use when `--data` is a `.zarr` SpatialData store.
`--outputDir DIR`	Where to write the result. Omit it and the input file is updated in place.
`--streamData`	A flag, not a value. Runs the step out of core — see Streaming.
`--verbose`	A flag. Off by default on the command line, unlike the Python API where `verbose=True` is usually the default.
`--label NAME`	The `obs`/`uns`/`layers` key to write results under.
`--maxWorkers N`	Caps parallelism.

Flag names match the Python parameter names exactly, camelCase and all, so imageId= in Python is --imageId on the command line.

A pipeline in bash¶

#!/usr/bin/env bash
set -euo pipefail

H5AD=big.h5ad

python -m scimappro.pp.rescale             --data "$H5AD" --gate manual_gates.csv --streamData
python -m scimappro.tl.phenotype           --data "$H5AD" --phenotype phenotype_workflow.csv --streamData
python -m scimappro.tl.spatial_distance    --data "$H5AD" --phenotype phenotype --streamData
python -m scimappro.tl.spatial_cooccurrence --data "$H5AD" --phenotype phenotype --streamData

Each step edits big.h5ad in place, so there are no intermediate files to manage.

Module names use the old spelling in a few places

spatial_distance, spatial_cooccurrence, and spatial_aggregate kept their snake_case module filenames from the port even though the functions they export are spatialDistance, spatialCooccurrence, and spatialAggregate. The table above is the authoritative mapping.