sort -z equivalent in Solaris ( support sorting of multi-line )

Question

I'm looking for "sort -z" (supported only in GNU) equivalent in Solaris, so I could have multi-line "block" support. I want to sort some text by chronological order ( date then time ), and that the sort will not break the text blocks in the process.

Example, I have:

2020-05-15:02:00:00:000 INFO[LF]
2020-05-15:02:00:02:000 INFO[LF]
[HT]some extra data as second line[LF]
[HT]2020-05-15:02:00:01:000 INFO[LF]
[HT]some extra data as second line[LF]
[HT]some extra data as third line[LF]
2020-05-15:02:00:04:000 INFO[LF]
2020-05-15:02:00:03:000 INFO[LF]

• [HT] is horizontal tab key and [LF] is \n

and I wish it to be sorted into

2020-05-15:02:00:00:000 INFO[LF]
2020-05-15:02:00:01:000 INFO[LF]
[HT]extra data as second line[LF]
[HT]some extra data as third line[LF]
2020-05-15:02:00:02:000 INFO[LF]
[HT]extra data as second line[LF]
2020-05-15:02:00:03:000 INFO[LF]
2020-05-15:02:00:04:000 INFO[LF]

on Centos/ubuntu (GNU) I manage to do that with: sort -zt : -k2,2 -k3,3n -k4,4n -k5,5n -k6,6n ( although not sure the milisec is being sorted ok ).

Please note:

1. I can't use msort - awk / sed / perl one-liner solutions are welcome.
2. each line is ending with LF, multi-line or not.
3. Multi lines ( 2nd line including and ahead ) starts with HF "tab" char.

Thanks :-)

Answer 1

With perl:

perl -0777 -ne 'print sort /^\d.*\n(?:\D.*\n)*/gm' your-file

That loads the whole file in memory (slurp mode with -0777 -p), extracts the blocks based on that regexps (line starting (^) with a digit (\d) following 0 or more (*) lines starting with a non-digit (\D)) and prints them after sorting.

Answer 2

To sort null-separated records, if there's a character that's guaranteed not to appear in your data, you can use tr to swap it with null bytes. However that won't help you directly since your data is separated by newlines, and even GNU sort doesn't have an option to treat only certain newlines as separators. You need to translate newline-tab sequences to something that doesn't contain a newline, then sort, then undo the original translation.

Assuming that your data doesn't contain tabs other than a single tab at the beginning of some lines, the translation can be to remove newlines before tabs, and add them back after sorting.

awk '/^\t/ {line = line $0; next}
     {print line; line = $0}
     END {print line}' |
sort |
awk '{gsub(/\t/, "\n\t"); print}'