Repeat an analysis¶

Use ordinary shell or workflow tools to run several complete projects. There is no additional batch command to learn: each project declares the model and analysis it needs.

Separate output folders¶

For a shell with the input folders projects/sample-a and projects/sample-b:

set -e
mkdir -p logs runs

for project in projects/sample-a projects/sample-b; do
    case_name=$(basename "$project")
    edi fit "$project" \
        --output "runs/$case_name" \
        --report machine \
        > "logs/$case_name.txt" \
        2> "logs/$case_name-errors.txt"
done

set -e
mkdir -p logs runs

for project in projects/sample-a projects/sample-b; do
    case_name=$(basename "$project")
    python -m edi fit "$project" \
        --output "runs/$case_name" \
        --report machine \
        > "logs/$case_name.txt" \
        2> "logs/$case_name-errors.txt"
done

This loop is a shell example, not a new edi API. It stops on a failed command and keeps reports separate from project output. Give each input a distinct output name and retain its original files.

Run without keeping model changes¶

Add --dry to execute fits without writing project changes. A dry run does real numerical work. In scan mode it keeps no resumable progress, so use an ordinary output folder for a scan you intend to continue.

A scan is a project workflow¶

If the measured files belong to one sequential or independent scan, declare that scan in the project instead of treating every pattern as an unrelated batch project.

• Sequential: the previous fit supplies the next pattern's initial parameters.
• Independent: every pattern starts from the same initial parameters.

Both process files serially. See shared fitting modes.

Keep failure information¶

Check the process exit status and retain stderr. A machine report is a versioned record; scripts that parse it should handle its schema and failed-run records explicitly. Do not treat the existence of a redirected output file as proof of a successful fit.

Save reports after fitting¶

python -m edi report runs/sample-a
python -m edi report runs/sample-a --check

Use the report guide to choose contents and check freshness.