--output_dir

Switch

--output_dir DIRECTORY

Description

Folder to write output files (plots, tagcloud data, correlation matrices, etc.) into, when you haven't given an explicit filename via --output_name. Use one or the other, not both.

Argument and Default Value

A directory path. Default: current directory (.), via dlac.DEF_OUTPUT_DIR.

Details

When --output_name is not given, makeOutputFilename() builds a filename by joining --output_dir with pieces describing the run (feature table, outcomes, controls, etc.) — e.g. ./out_test/feat.1to3gram.msgs100u.user_id.age.

Known issue: for the --correlate + --rmatrix/--csv output path specifically, dlatkInterface.py then wraps that already-prefixed filename in a second os.path.join(args.outputdir, ...) call, doubling the directory (e.g. ./out_test/./out_test/....csv) and crashing with FileNotFoundError since that doubled path doesn't exist. Verified by running --correlate --csv --output_dir ./out_test with no --output_name. Workaround: always pass --output_name explicitly when using --correlate with --csv/--rmatrix — that path bypasses --output_dir entirely and works correctly.

Other Switches

Optional Switches:

--output_name (takes precedence over this switch if both are given)

Example Commands

# Broken as of this writing (see Known issue above):
dlatkInterface.py -d dla_tutorial -t msgs -c user_id -f 'feat$1gram$msgs$user_id$16to16' \
--outcome_table blog_outcomes --outcomes age --correlate --csv --output_dir ./results

# Works: use --output_name instead
dlatkInterface.py -d dla_tutorial -t msgs -c user_id -f 'feat$1gram$msgs$user_id$16to16' \
--outcome_table blog_outcomes --outcomes age --correlate --csv --output_name ./results/corr