Renaming and File Formats

Two groups of options that deal with the shape of files: renaming the columns of the masked dataset (--rename-columns, --rename-except, --rename-prefix), and reading files that are not a plain comma-separated table with a header (--delimiter, --no-header, --columns, --comment-chars, --output-name).

1. Renaming columns

Column names can give away what a column means (HIVVL, SURGDATE). Renaming hides that. Every rename is recorded in keymap.txt (section 1), so reverse puts the original names back.

Automatic names

data-obfuscator run --input pk_study.csv --output ./masked --level 2 --seed 2026 --rename-columns WT,AGE,SEX

WT, AGE and SEX are the 13th, 14th and 15th columns, so they become COL13, COL14 and COL15. Note that renaming only changes the header: the values are masked exactly as they would have been (SEX is still recoded).

Names you choose

data-obfuscator run --input pk_study.csv --output ./masked --level 2 --seed 2026 --rename-columns WT=BODYSIZE,AGE=YEARS

Everything except a list

data-obfuscator run --input pk_study.csv --output ./masked --level 2 --seed 2026 \
  --rename-except ID,TIME,AMT,DV,EVID,MDV,CMT --rename-prefix X

The seven columns listed keep their names (the ones a NONMEM control stream needs); all the others become X02, X03, X04, ... In PowerShell put the list in quotes.

Renames are recorded and undone

reverse undoes renaming as its first step (rename: BODYSIZE -> WT), and the restored header is the original one. A column that the key does not know (one the LLM added) is kept and placed after the restored columns.

Column names are protected data too: summary.txt, column_report.txt and config.used.txt keep the original names (and so does the key), so they must stay private.

2. Reading other file formats

The program reads CSV, TSV, semicolon-separated and whitespace-separated files. The delimiter is detected from the header line; say so when detection cannot work.

--delimiter accepts auto, comma, tab, semicolon and space (also csv, tsv, whitespace, nonmem).

A file without a header

A NONMEM data file usually has no header. Say so and give the names:

data-obfuscator run --input pk_whitespace.dat --output ./masked --level 1 --seed 2026 \
  --no-header --columns ID,TIME,AMT,DV,EVID,MDV,CMT,WT,AGE

The input has nine columns and no names; the output is obfuscated.dat (the extension follows the input), with the header ID TIME AMT DV EVID MDV CMT WT AGE added and the subject numbers recoded. If your NONMEM $DATA statement expects no header, add IGNORE=@ or remove the first line.

Comment lines

A file exported with notes at the top works as it is:

# exported 2026-10-12 by the data office
; do not edit
ID,USUBJID,STUDYID,SITEID,ARM,TIME,...

The two comment lines are skipped when reading. They are not written to the output. To use other comment characters, for example !: --comment-chars '!'.

Rows with the wrong number of fields

If a data row has more or fewer fields than the header, the program warns, because NONMEM would reject the file:

[WARN] 1 data row(s) do not have 22 fields like the header - NONMEM would reject this file. They keep their shape in the
       masked copy so the problem stays visible; fields beyond the header are copied UNMASKED.

Read this warning carefully: a field beyond the header is not masked because the program does not know what it is. Fix the file (usually an unquoted comma inside a text field) and run again.

3. The name of the masked file

--output-name study_masked.csv writes the masked dataset under that name instead of obfuscated.csv. The other files keep their names.

data-obfuscator run --input pk_study.csv --output ./masked --level 1 --seed 2026 --output-name study_masked.csv

Next

Options Files shows how to keep all of these options in a file.