Use '+' as alternative for 'then'?

[zwets]

Only recently discovered miller and wow does it fill a gap! It's now the monkey wrench in my toolbox, giving well-worn grep, sed, cut and awk a rest. Thanks so much for conceiving and developing this!

My suggestion would be to offer + as an alternative for then. The plus symbol has no meaning for shells, conveys some of the 'pipe' or 'then' meaning, and above all provides a clear visual cue of separating two sides of a pipeline - including suggesting (correctly) that it has lowest binding priority.

I'm well aware that this comes down to personal taste, so feel free to drop the issue.

[zwets]

PS: my one peeve with mlr is that -t isn't the implicit default, though surely this is due to personal trauma. However short, having to type it every time reminds of the pain of awk -F '\t' 'begin {OFS=FS}' and of sort and join's dreaded "multi-character tab ‘\\t’" unless you recall that magic incantation -t $'\t' and aren't on sh.

Yes, scrapping that compulsory -t would bring a disproportionate improvement to my quality of life.

[johnkerl]

@zwets have you taken a look at https://miller.readthedocs.io/en/latest/customization/ ?

[johnkerl]

@zwets + is totally doable! :)

[zwets]

@zwets have you taken a look at https://miller.readthedocs.io/en/latest/customization/ ?

Sure, and I would put it in ~/.mlrrc if it weren't for portability. My scripts will often be run by others (on our compute cluster) so I'm wary of using dotfile settings or aliases. Great for offhand work, but then you need to always remember to be formal in scripts. If it weren't for this, I'd long had alias awk='awk -F ... in my ~/.bash_aliases. :-)

I could drop a wrapper tmlr in the clusterwide /bin, just like I've done for tawk, tjoin and ttsort (oww, nameclash with tsort in coreutils!), but that's too heavy-handed for just adding that -t.

It's just that - at least in bioinformatics - tab-separated has pretty much become the default, and mlr looks like just the tool to succeed the golden oldies (awk, grep, sed, ...) for handling it. My thinking was that mlr is still young enough to get this default right, without even breaking backward compatibility, but I overlooked that it already has the DKVP default! A somewhat unconventional choice, but I can learn to live with that -t.

[johnkerl]

@zwets :)

The DKVP default goes WAY back ... to when it was just me. I'd be happy to drop that and use CSV as the default, since (as last year's survey showed) CSV is the format Miller users use most (although all formats get used).

mlr -t as default ... I'm sure is the right thing for a very large demographic but it appears the mlr -c crowd is bigger still ...

[zwets]

mlr -t as default ... I'm sure is the right thing for a very large demographic but it appears the mlr -c crowd is bigger still ...

Hmm, well, ok then. ;-) At least I'll have happier trainees. Those on Windows take forever to figure out how to open a TSV in Excel - while they just saw me double-click it and it opened in LibreOffice.

They'll be só happy that bioinformatics tools can actually produce "Excel Files" - just that I'll now have to teach them mlr too. :p

[johnkerl]

@zwets apologies on my delay in responding to this (and to everything else Miller-related) -- the last few months I've self-saturated at work on https://github.com/single-cell-data/TileDB-SingleCell/tree/main/apis/python.

#1049 implements + as an alias for then.