gawk vs mawk: case‑insensitive matching compatibility

gawk provides case‑insensitive matching via the global variable IGNORECASE or the inline flag (?i), while mawk and nawk do not implement either feature, making scripts non‑portable.

Likely explanation

The awk language specification does not define a portable case‑insensitive mechanism; implementations diverge. gawk treats IGNORECASE as a boolean that affects all regexes in the program, and (?i) toggles case‑insensitivity for the surrounding pattern. mawk and nawk ignore these, so a pattern that matches case‑insensitively in gawk may fail or match differently in mawk/nawk.

Confirmed facts

• GNU awk (gawk) recognizes the IGNORECASE variable and the (?i) inline flag.
• mawk and nawk lack both IGNORECASE and (?i), resulting in case‑sensitive matching by default.
• POSIX awk does not mandate case‑insensitive matching, so portable scripts must avoid implementation‑specific features.
• Case‑folding can be achieved by converting both pattern and subject to a common case, e.g., using tolower() on the subject and lower‑casing the pattern.
• Setting LC_ALL=C ensures locale‑independent lower‑casing.

Best‑practice emulation

1. Normalize the subject string to lower case: subject = tolower(subject) (or use tolower($0) in the rule).
2. Lower‑case the pattern: pattern = tolower(pattern) before applying it.
3. Use a single case‑insensitive regex that does not rely on IGNORECASE or (?i), e.g., /pattern/ where pattern is already lower‑cased.
4. Optionally wrap the whole script in a portable shim: BEGIN { IGNORECASE=1 } for gawk only, guarded by a version check.

CI verification steps

• Create a minimal test file test.awk containing a mixed‑case pattern and a sample line.
• Run the script with gawk, mawk, and nawk and capture exit status and stdout.
• Assert that the output strings are identical; any difference signals non‑portable behavior.
• In CI, automate the above commands and fail the job if the captured outputs diverge.

Missing diagnostic

Confirm whether the scripts set IGNORECASE globally or rely on inline (?i), as this determines if a simple case‑normalisation shim will suffice.