Quickie - Regexp for a string not at the beginning of the line

[Rivka Miller]

Why you just don't give us the string/input, say a line or two, and what you want off of it, so we can tell better what to suggest

no one has really helped yet.

I want to search and modify.

I dont wanna be tied to a specific language etc so I just want a
regexp and as many versions as possible. Maybe I should try in emacs
and so I am now posting to emacs groups also, although javascript has
rich set of regexp facilities.

examples

$hello$ should not be selected but
not hello but all of the $hello$ and $hello$ ... $hello$ each one
selected

=================
original post
=================


Hello Programmers,

I am looking for a regexp for a string not at the beginning of the
line.

For example, I want to find $hello$ that does not occur at the
beginning of the string, ie all $hello$ that exclude ^$hello$.

In addition, if you have a more difficult problem along the same
lines, I would appreciate it. For a single character, eg < not at the
beginning of the line, it is easier, ie

^[^<]+<

but I cant use the same method for more than one character string as
permutation is present and probably for more than one occurrence,
greedy or non-greedy version of [^<]+ would pick first or last but not
the middle ones, unless I break the line as I go and use the non-
greedy version of +. I do have the non-greedy version available, but
what if I didnt?

If you cannot solve the problem completely, just give me a quick
solution with the first non beginning of the line and I will go from
there as I need it in a hurry.

Thanks

[Ed Morton]

Post by Rivka Miller

Why you just don't give us the string/input, say a line or two, and what you want off of it, so we can tell better what to suggest

no one has really helped yet.

Because there is no solution - there IS no _RE_ that will match a string not at
the beginning of a line.

Now if you want to know how to extract a string that matches an RE in awk,
that'd be (just one way):

awk 'match($0,/.[$]hello[$]/) { print substr($0,RSTART+1,RLENGTH-1) }'

and other tools would have their ways of producing the same output, but that's
not the question you're asking.

Ed.

Post by Rivka Miller
I want to search and modify.
I dont wanna be tied to a specific language etc so I just want a
regexp and as many versions as possible. Maybe I should try in emacs
and so I am now posting to emacs groups also, although javascript has
rich set of regexp facilities.
examples
$hello$ should not be selected but
not hello but all of the $hello$ and $hello$ ... $hello$ each one
selected
=================
original post
=================
Hello Programmers,
I am looking for a regexp for a string not at the beginning of the
line.
For example, I want to find $hello$ that does not occur at the
beginning of the string, ie all $hello$ that exclude ^$hello$.
In addition, if you have a more difficult problem along the same
lines, I would appreciate it. For a single character, eg < not at the
beginning of the line, it is easier, ie
^[^<]+<
but I cant use the same method for more than one character string as
permutation is present and probably for more than one occurrence,
greedy or non-greedy version of [^<]+ would pick first or last but not
the middle ones, unless I break the line as I go and use the non-
greedy version of +. I do have the non-greedy version available, but
what if I didnt?
If you cannot solve the problem completely, just give me a quick
solution with the first non beginning of the line and I will go from
there as I need it in a hurry.
Thanks

[Ben Bacarisse]

Post by Rivka Miller

Why you just don't give us the string/input, say a line or two, and
what you want off of it, so we can tell better what to suggest

no one has really helped yet.

Really? I was going to reply but then I saw Janis had given you the
answer. If it's not the answer, you should just reply saying what it is
that's wrong with it.

Post by Rivka Miller
I want to search and modify.

Ah. That was missing from the original post. You can't expect people
to help with questions that weren't asked! To replace you will usually
have to capture the single preceding character. E.g. in sed:

sed -e 's/\(.\)$hello\$/\1XXX/'

but some RE engines (Perl's, for example) allow you specify zero-width
assertions. You could, in Perl, write

s/(?<=.)\$hello\$/XXX/

without having to capture whatever preceded the target string. But
since Perl also has negative zero-width look-behind you can code your
request even more directly:

s/(?<!^)\$hello\$/XXX/

Post by Rivka Miller
I dont wanna be tied to a specific language etc so I just want a
regexp and as many versions as possible. Maybe I should try in emacs
and so I am now posting to emacs groups also, although javascript has
rich set of regexp facilities.

You can't always have a universal solution because different PE
implementations have different syntax and semantics, but you should be
able to translate Janis's solution of matching *something* before your
target into every RE implementation around.

Post by Rivka Miller
examples
$hello$ should not be selected but
not hello but all of the $hello$ and $hello$ ... $hello$ each one
selected

I have taken your $s to be literal. That's not 100 obvious since $ is a
common (universal?) RE meta-character.

<snip>

[Rivka Miller]

Thanks everyone, esp this gentleman.

The solution that worked best for me is just to use a DOT before the
string as the one at the beginning of the line did not have any char
before it. I guess, this requires the ability to ignore the CARAT as
the beginning of the line.

I am a satisfied custormer. No need for returns. :)

Post by Rivka Miller

Why you just don't give us the string/input, say a line or two, and
what you want off of it, so we can tell better what to suggest

no one has really helped yet.

Really?  I was going to reply but then I saw Janis had given you the
answer.  If it's not the answer, you should just reply saying what it is
that's wrong with it.

Post by Rivka Miller
I want to search and modify.

Ah.  That was missing from the original post.  You can't expect people
to help with questions that weren't asked!  To replace you will usually
  sed -e 's/\(.\)$hello\$/\1XXX/'
but some RE engines (Perl's, for example) allow you specify zero-width
assertions.  You could, in Perl, write
  s/(?<=.)\$hello\$/XXX/
without having to capture whatever preceded the target string.  But
since Perl also has negative zero-width look-behind you can code your
  s/(?<!^)\$hello\$/XXX/

Post by Rivka Miller
I dont wanna be tied to a specific language etc so I just want a
regexp and as many versions as possible. Maybe I should try in emacs
and so I am now posting to emacs groups also, although javascript has
rich set of regexp facilities.

You can't always have a universal solution because different PE
implementations have different syntax and semantics, but you should be
able to translate Janis's solution of matching *something* before your
target into every RE implementation around.

Post by Rivka Miller
examples
$hello$ should not be selected but
not hello but all of the $hello$ and $hello$ ... $hello$ each one
selected

I have taken your $s to be literal.  That's not 100 obvious since $ is a
common (universal?) RE meta-character.
<snip>
--
Ben.

[Janis Papanagnou]

Post by Rivka Miller
Thanks everyone, esp this gentleman.

Who is "this"?

Post by Rivka Miller
The solution that worked best for me is just to use a DOT before the
string as the one at the beginning of the line did not have any char
before it.

Which was what I suggested, and where you rudely answered...

Post by Rivka Miller
no one has really helped yet.

And obviously...

Post by Rivka Miller
I am a satisfied custormer.

...your perception about yourself and about the role of us
Usenet posters seems also not be very sane. Good luck.

[Ben Bacarisse]

Post by Rivka Miller
Thanks everyone, esp this gentleman.

Kind of you to single me out, but it was Janis Papanagnou who first
posted the solution that you say "works best" for you.

<snip>

[Ed Morton]

Post by Rivka Miller
Thanks everyone, esp this gentleman.
The solution that worked best for me is just to use a DOT before the
string as the one at the beginning of the line did not have any char
before it.

That's fine but do you understand that that is not an RE that matches on
"$hello$ not at the start of a line", it's an RE that matches on "<any
char>$hello$ anywhere in the line"? There's a difference - if you use a tool
that prints the text that matches an RE then the output if the first RE existed
would be "$hello$" while the output for the second RE would be "X$hello$" or
"Y$hello$" or....

In some tools you can use /(.)$hello$/ or similar to ignore the first part of
the RE "(.)" and just print the second "$hello", but that ability and it's
syntax is tool-specific, you still can't say "here's an RE that does this",
you've got to say "here's how to find this text using tool <whatever>".

Ed.

Post by Rivka Miller
I guess, this requires the ability to ignore the CARAT as the beginning of the line.
I am a satisfied custormer. No need for returns. :)

Post by Ben Bacarisse

Post by Rivka Miller

Why you just don't give us the string/input, say a line or two, and
what you want off of it, so we can tell better what to suggest

no one has really helped yet.

Really? I was going to reply but then I saw Janis had given you the
answer. If it's not the answer, you should just reply saying what it is
that's wrong with it.

Post by Rivka Miller
I want to search and modify.

Ah. That was missing from the original post. You can't expect people
to help with questions that weren't asked! To replace you will usually
sed -e 's/\(.\)$hello\$/\1XXX/'
but some RE engines (Perl's, for example) allow you specify zero-width
assertions. You could, in Perl, write
s/(?<=.)\$hello\$/XXX/
without having to capture whatever preceded the target string. But
since Perl also has negative zero-width look-behind you can code your
s/(?<!^)\$hello\$/XXX/

Post by Rivka Miller
I dont wanna be tied to a specific language etc so I just want a
regexp and as many versions as possible. Maybe I should try in emacs
and so I am now posting to emacs groups also, although javascript has
rich set of regexp facilities.

You can't always have a universal solution because different PE
implementations have different syntax and semantics, but you should be
able to translate Janis's solution of matching *something* before your
target into every RE implementation around.

Post by Rivka Miller
examples
$hello$ should not be selected but
not hello but all of the $hello$ and $hello$ ... $hello$ each one
selected

I have taken your $s to be literal. That's not 100 obvious since $ is a
common (universal?) RE meta-character.
<snip>
--
Ben.

[a***@anon.anon]

On Thu, 25 Oct 2012 18:08:53 -0700 (PDT), Rivka Miller

Post by Rivka Miller
no one has really helped yet.

We regret that you are not a satisfied customer.

Please take your receipt to the cashier and you will receive double your
money back according to our "you must be satisfied" guarantee.

[l***@gmail.com]

I'm still learning Emacs, and don't yet know how to build new customized functionality using Elisp, but I would guess it wouldn't be too hard to code something like this, would my guess be right? I'm visualizing some little piece of Lisp code that would maybe search for the string in question and each time it finds an occurrence, you could (or could you?) check the position of the found instance on the line and if it's in the first position, pretend it wasn't really found and keep going...can a person do that kind of thing with Emacs to solve this fellow's problem?

[Axel Wegen]

Post by l***@gmail.com
I'm visualizing some little piece of Lisp code that would maybe search
for the string in question and each time it finds an occurrence, you
could (or could you?) check the position of the found instance on the
line and if it's in the first position, pretend it wasn't really found
and keep going...can a person do that kind of thing with Emacs to
solve this fellow's problem?

Here is some Emacs Lisp that I think does what you described:

(defun foo (re)
(interactive "sRE: ")
(while (and (re-search-forward re nil t)
(= (- (match-beginning 0) (point-at-bol)) 0))
nil
)
)

Being a novice at programming and Emacs Lisp that might not be the
smartest way, not even a particularly good one, to write such a function.

Another possibility is to put '.+' in front of a regular expression to
exclude matched-strings at the beginning of a line.

--
Axel Wegen