What is ~(X) pattern option

[phidebian]

While ramping up my knowledge (for bug-833) of pattern in =~ construct I bump into this on the ksh man page
`
~(option:pattern-list)
...

          X      The remainder of the pattern uses augmented  regular  ex‐
                 pression  syntax  like the -X option of the AT&T AST ver‐
                 sion of the grep(1) command.

`

Where could I find the ast-grep man page beside the well known abstract syntax grep man page, i.e where is the AT&T man page.

[dannyweldon]

You can find the man page here, but it doesn't provide any further explanation, which I imagine was intended to eventually be forthcoming:

https://web.archive.org/web/20151120075529/http://www2.research.att.com/~astopen/man/man1/egrep.html

According to regex.h:

ksh/src/lib/libast/include/regex.h

Line 43 in af11607

#define REG_AUGMENTED 0x00000001 /* enable ! & < > */

And compared to extended mode:

ksh/src/lib/libast/include/regex.h

Line 44 in af11607

#define REG_EXTENDED 0x00000002 /* enable ( | ) */

It would be worthwhile to verify that it works and provide a similar comment on the ksh man page itself.

[phidebian]

OK I think I got it, the ~(X) is for augmented regular expression known as ARE, this is like ~(E) extended regular expression known as ERE but with the more RE meta char has noted by @dannyweldon with /* enable ! & < > */

So it bring things like

$ x=111222333444

$ echo ${x//~(X)(1)|(2)/A}
AAAAAA333444

$ echo ${x//~(X)(1)!(2)/A}
A333444

[dannyweldon]

@phidebian, can you explain what this is doing with the ! character? I understand that the ones and twos are all being replaced with an A, but what does the ! syntax mean exactly, as I can replace the first sub-expression with any single character and the replacement still works?

$ echo ${x//~(X)(1)!(2)/A}
A333444

Also, it looks like the < and > characters are for matching the beginning and endings of word boundaries, a bit like \b:
/* enable ! & < > */
I have not been able to figure out how the & is meant to be used.

[phidebian]

Frankly I think it is doing magic. Yes < and > seems word anchor but can't prove this beside a test

    TC$ x='123 234 345 456 222 444 666'
    
    TC$  echo ${x//~(X)<2/A}
    123 A34 345 456 AAA 444 666
    
    TC$ echo ${x//~(X)2>/A}
    123 234 345 456 22A 444 666
    
    TC$ echo ${x//~(X)4>/A}
    123 23A 345 456 222 44A 666

See here <2 seems t say match all 2 from a word start
2> seem to say 2 at end of word, yet there is no 'all' 2 at word end so < and > seems different.
ditto for 4> only 1 4 at end of word
May be <4> means all 4 at words start up to end of word

TC$ echo ${x//~(X)<4>/A}
123 234 345 456 222 444 666

Dunno how to use this...

Amazingly tests/comvario.sh reference only the & meta char in an non-sensical way to me, if we remove the ~(X) alltogether, we would remove the tests as well.