BEGINFILE/ENDFILE vs. FNR==1: Choosing Per‑File State Logic in Awk

Goal

Implement per‑file counters and summaries in an awk script that processes many input files.

Constraint

The script must run on a variety of Unix-like systems where the default awk may be POSIX awk, mawk, BusyBox awk, or gawk. The choice of per‑file logic affects portability, readability, and error handling.

Uncertainty

While the FNR==1 idiom works everywhere, it requires explicit checks and can be error‑prone with empty or unreadable files. The gawk extensions BEGINFILE and ENDFILE provide cleaner hooks but lock the script to gawk and have subtle semantics for empty or failed files.

Specific Questions

• Does the target deployment environment guarantee that awk is gawk, or could it be a POSIX or minimal awk?
• Will the input set include empty files or files that might fail to open, and how should those cases be handled?
• Is the added maintenance overhead of the FNR==1 idiom acceptable compared to the clarity of BEGINFILE/ENDFILE?