csv_examples

CSV output for wrdata (Enhancement-231)

ngspice's wrdata writes whitespace-separated columns, with every vector prefixed by its own scale column. The available knobs (wr_singlescale, wr_vecnames, wr_onespace, numdgt) never produced a real comma-separated file, so exporting to a spreadsheet or pandas/csv reader meant a post-processing step. This enhancement gives wrdata a first-class CSV mode.

Usage

Two equivalent ways — a sticky option and a per-call flag:

* option form
set wr_csv
wrdata out.csv v(out) i(v1)
unset wr_csv

* flag form (any position; enables CSV for just this write)
wrdata -csv out.csv v(out) i(v1)
wrdata out.csv v(out) i(v1) -csv

CSV mode writes:

• a header row of names (implies wr_vecnames),
• a single shared scale column — time, frequency, … — written once (implies wr_singlescale, so all vectors must share one scale length), and
• comma-separated values with no leading/trailing padding.

time,v(in),v(out)
0.00000000e+00,0.00000000e+00,0.00000000e+00
1.00000000e-09,1.00000000e-03,9.99000999e-07
...

For .ac, each complex vector becomes two columns (real, imaginary) under its name. Precision follows set numdgt (default 8 significant digits).

The -csv flag is a per-call alias for set wr_csv: it enables CSV for that one write and restores the prior global state afterward, so it never leaks into later wrdata calls. Default (no-CSV) output is byte-for-byte unchanged.

Verify

python3 verify_csv.py

Checks that default output is unmodified; that set wr_csv and wrdata -csv (in first / middle / last argument positions) produce identical, header-topped, comma-separated files; that the numbers match the default format bit-for-bit; that .ac emits real/imag columns and .tran a time scale; and that the flag does not leak into the following plain wrdata.