Use from Python

The simplest Python integration is to call the CLI with --json and parse the result with the standard json module.

Why this is the recommended path

• no Perl-to-Python bridge is required
• the validator keeps one stable machine-readable output format
• the same command works in scripts, notebooks, and pipelines

Minimal example

import json
import subprocess

completed = subprocess.run(
    [
        "bin/omop-csv-validator",
        "--ddl", "ddl/OMOPCDM_postgresql_5.4_ddl.sql",
        "--input", "example/DRUG_EXPOSURE.csv",
        "--json",
    ],
    capture_output=True,
    text=True,
    check=False,
)

result = json.loads(completed.stdout)
result

What you get back

The JSON result contains:

• input_file
• schema_name
• ok
• error_count
• row_errors

If validation cannot even start, such as when no schema can be inferred, the result also contains fatal_error.

These fields are intended to be the stable automation interface for Python clients.

Typical Python branching

if result["ok"]:
    print("CSV validated successfully")
elif "fatal_error" in result and result["fatal_error"] is not None:
    raise RuntimeError(result["fatal_error"])
else:
    print(result["row_errors"])

Batch validation in Python

For multiple OMOP CSV files, loop over files and run the validator once per file.

import glob
import json
import subprocess

results = []

for csv_file in glob.glob("exports/*.csv"):
    completed = subprocess.run(
        [
            "bin/omop-csv-validator",
            "--ddl", "ddl/OMOPCDM_postgresql_5.4_ddl.sql",
            "--input", csv_file,
            "--json",
        ],
        capture_output=True,
        text=True,
        check=False,
    )

    results.append(json.loads(completed.stdout))

That pattern keeps the validator simple while giving Python users a straightforward batch workflow.

Separator override is usually unnecessary

The CLI normally infers the separator. Add --sep in Python only if you know the file needs an explicit override.

Row numbering

Row numbers in row_errors are data rows, not physical line numbers including the header. The first row after the header is row 1.