More RS/ASCII updates: clearer update method

[johnkerl]

Implements #1618 (reply in thread)

[GlassGruber]

Nice! But isn't the deprecated old values mapping missing?
Something like this?

    "ascii_esc":       ASCII_ESC,
    "ascii_etx":       "\\x04",     // deprecated
    "ascii_etx_char":  ASCII_ETX,
    "ascii_fs":        ASCII_FS,
    "ascii_gs":        ASCII_GS,
    "ascii_null":      "\\x01",     // deprecated
    "ascii_null_char": ASCII_NULL,
    "ascii_rs":        ASCII_RS,
    "ascii_soh":       "\\x02",     // deprecated
    "ascii_soh_char":  ASCII_SOH,
    "ascii_stx":       "\\x03",     // deprecated
    "ascii_stx_char":  ASCII_STX,

Sorry I should make a pr but I'm using another machine and it's a bit cumbersome from this one

[johnkerl]

@GlassGruber TBH I'd rather surface a hard error if someone uses an old mapping that's wrong -- then they can fix their pipelines ...

[johnkerl]

Maybe it can print a warning for the next release, then, old syntax goes away in the release after? I.e. a deprecation cycle ...

I definitely do not want to keep misnamed values around forever ..... :|

[GlassGruber]

Totally makes sense!

[balki]

I think for this case, it is better to break backwards compatibility and highlight clearly in ReleaseNotes. It is very unlikely for someone to use ascii_null to mean SOH. I think users would have realized the aliases don't work and replaced with explicit forms like \0 or \x00.

Renaming makes them verbose (extra _char) and inconsistent naming as some aliases don't have _char while some have.

Also the old name does not cause miller to fail but take them literally.

miller on  ascii:kerl/ascii-more via 🐹 v1.23.1 
❯ echo "foo,bar\n1,2\n3,4" |  go run cmd/mlr/main.go --c2d --ofs ascii_null cat | od -c
0000000   f   o   o   =   1   a   s   c   i   i   _   n   u   l   l   b
0000020   a   r   =   2  \n   f   o   o   =   3   a   s   c   i   i   _
0000040   n   u   l   l   b   a   r   =   4  \n
0000052

[johnkerl]

@balki I like that! @GlassGruber what do you think?

[GlassGruber]

I think it makes sense, especially this

It is very unlikely for someone to use ascii_null to mean SOH. I think users would have realized the aliases don't work and replaced with explicit forms like \0 or \x00.

it's just what I did!
I'm usually for taking it slow about breaking compatibility, but yea no much sense in keeping things as in current state. Also removing that extra _char is a nice +1!

[johnkerl]

Thanks @balki @GlassGruber -- I'll close this & tag 6.13.0.