Query expressions

Used by subset --obs-query and --var-query.

Operator	Example
==, !=	cluster == Cortex_2
<, <=, >, >=	n_counts > 1000
in, not in	cluster in A,B,C
and, or, not	n_counts > 500 and not cluster == Fiber_tract
parentheses	(a == 1 or b == 2) and c > 3

Values with spaces need quoting: label == "cell type A". Equality and membership compare the column’s string form; the ordering operators parse it as a number, and entries that will not parse simply do not match.

Only the columns a query mentions are read, so filtering on one annotation does not touch the rest of the frame.

This is deliberately not SQL. For anything more involved — joins, aggregates, window functions — export to CSV and use duckdb:

adata export dataframe data.h5ad obs -o cells.csv
duckdb -noheader -list -c \
  "SELECT _index FROM 'cells.csv' WHERE cluster='Cortex_2' AND n_counts > 1000" \
  > barcodes.txt
adata subset data.h5ad -o cortex.h5ad --obs barcodes.txt