mcrun: only take values made of numbers as scan syntax

[tkittel]

After iterations with @willend, me and our coding slave Claude, this now fixes yet another place where one could not easily pass through NCrystal cfg-strings.

I did review the changes - and claude wrote the rest here

mcrun took every parameter value containing : or , as scan syntax. So NCrystal cfg-strings (and e.g. Windows paths) with such characters could not be given on the command line:

mcrun Ge.instr "cfg=Ge_sg227.ncmat;dir1=@crys_hkl:5,1,1@lab:0,0,1;dir2=@crys_hkl:0,-1,1@lab:0,1,0"
CRITICAL: Parameter "cfg" uses "a:delta:b" syntax but has 5 colon-separated part(s) ...

Values with :: (stdlib::...) skipped the colon check. But they then went to the comma parsing, where a value with commas was dropped (and the instrument ran with the default).

Now only values made of numbers separated by : or , are scan syntax. Other values are used as they are. This replaces the special case for ::. With -L, values are still lists separated by commas, as before. A comma within a list entry can now be escaped as \,, which is mentioned in the -L help:

mcrun -L ... x=1,2 "cfg=Ge_sg227.ncmat;dir1=@crys_hkl:5\,1\,1@lab:0\,0\,1;...,Ge_sg227.ncmat;dir1=@crys_hkl:5\,1\,1@lab:0\,1\,0;..."

Tested by comparing with the current mcrun on 44 command lines:

• Identical results for all of these:
  • scans: -N ranges (also negative and exponent numbers), a:delta:b, -M (also with -N2,3 and mixed syntaxes), -L with lists of file names and cfg-strings (also with ::), -L with min:delta:max, --optimize, --seeds;
  • errors and empty values: x=1:2, x=1,2,, x=.
• Now work:
  • NCrystal cfg-strings, e.g. dir1=@crys_hkl:5,1,1...;
  • Windows paths like C:\data\a.dat, also in -L lists;
  • -L entries with \,;
  • f=a.dat,b.dat without -L, which is now that literal string instead of being dropped.
• Changed errors:
  • x=1:a:3 is now rejected by the instrument ("Invalid value '1:a:3' for floating point parameter x") instead of by mcrun;
  • x=1::3 is the other way round;
  • -N2 f=b.dat,c.dat now gives "No interval range specified" instead of a Python exception.
• A trailing comma in a non-numeric value (f=a.dat,) is still ignored with a warning, as before.

🤖 Generated with Claude Code

[willend]

Cool @tkittel, indeed great (but also busy) times with these new code-slaves. 🤖 🤣

I have a slurm batch somewhere that does a multi-dimensional scan with NCrystal strings, let me run that in addition to the PR tests - will merge if that also comes back healthy.