query-replace-regexp on a string with '=' sign

query-replace-regexp on a string with '=' sign

[Pankaj Jangid]

I have a line like this,

HELLO=WORLD

now I want to downcase the part before the equal (=) sign.

I am approaching like this

M-x query-replace-regexp RET \(.+\)= RET \,(downcase \1)= RET

But this doesn't work. It works if the equal sign is replaced with
another non-special ASCII character.

Where am I wrong?

Re: query-replace-regexp on a string with '=' sign

[Emanuel Berg]

Pankaj Jangid wrote:

> I have a line like this,
>
> HELLO=WORLD
>
> now I want to downcase the part before the equal (=) sign.

(while (re-search-forward "\\(.+\\)\\(=\\)" nil t)
  (replace-match
   (concat (downcase (match-string-no-properties 1))
           (match-string-no-properties 2) ) t))
;; We=
;; Are=
;; Robots=

-- 
underground experts united
https://dataswamp.org/~incal

RE: [External] : Re: query-replace-regexp on a string with '=' sign

[Drew Adams]

> > HELLO=WORLD
> > I want to downcase the part before the equal (=) sign.
> 
> (while (re-search-forward "\\(.+\\)\\(=\\)" nil t)...

In most cases you don't want the first group to match
any `=' chars, in which case it should be something
like \([^=]+\).

Re: [External] : Re: query-replace-regexp on a string with '=' sign

[Emanuel Berg]

Drew Adams wrote:

>>> HELLO=WORLD
>>> I want to downcase the part before the equal (=) sign.
>> 
>> (while (re-search-forward "\\(.+\\)\\(=\\)" nil t)...
>
> In most cases you don't want the first group to match any
> `=' chars, in which case it should be something like
> \([^=]+\).

I don't know why mine works, why not everything including the
equals sign _and_ the part after that ends up in the first
group?

But it does work, while replacing the RE with
"\\([^=]+\\)\\(=\\)" doesn't, it downcases the whole thing.

Here, try it yourself:

(while (re-search-forward "\\(.+\\)\\(=\\)" nil t)
  (replace-match
   (concat (downcase (match-string-no-properties 1))
           (match-string-no-properties 2) ) t))

;; We=WE
;; Are=ARE
;; Robots=ROBOTS

-- 
underground experts united
https://dataswamp.org/~incal

Re: query-replace-regexp on a string with '=' sign

[Tassilo Horn]

Pankaj Jangid <pankaj@codeisgreat.org> writes:

> I have a line like this,
>   HELLO=WORLD
> now I want to downcase the part before the equal (=) sign.
> I am approaching like this
>
>  M-x query-replace-regexp RET \(.+\)= RET \,(downcase \1)= RET
>
> But this doesn't work.

Where "doesn't work" means "doesn't downcase but keeps HELLO", right?
Have a look at (info "(emacs) Replacement and Lax Matches") which
explains it.  I think the default `case-replace' value of t is the
problem here.

That said, `case-replace' is actually a nice feature.  It's just that
with an explicit `downcase' it's very counter-intuitive.  I'd suggest to
report that as a bug.

What you can do is instead of saying 'y' when asked to replace the
current occurrence do 'E RET' instead.

> It works if the equal sign is replaced with another non-special ASCII
> character.

I don't understand what you mean here.  When I change the = in the
replacement to, say, the letter t, HELLO= gets replaced with HELLOT
which is the same behavior as above...

Bye,
Tassilo

Re: query-replace-regexp on a string with '=' sign

[Emanuel Berg]

Tassilo Horn wrote:

> That said, `case-replace' is actually a nice feature.
> It's just that with an explicit `downcase' it's very
> counter-intuitive. I'd suggest to report that as a bug.

Yeah, but does it work like that, explicit or not, isn't it
just an arbitrary function, namely `downcase'?

-- 
underground experts united
https://dataswamp.org/~incal

Re: query-replace-regexp on a string with '=' sign

[Pankaj Jangid]

Tassilo Horn <tsdh@gnu.org> writes:

>> I have a line like this,
>>   HELLO=WORLD
>> now I want to downcase the part before the equal (=) sign.
>> I am approaching like this
>>
>>  M-x query-replace-regexp RET \(.+\)= RET \,(downcase \1)= RET
>>
>> But this doesn't work.
>
> Where "doesn't work" means "doesn't downcase but keeps HELLO", right?

Yes. Suppose I have these lines in a region

HELLO=WORLD
HELLOXXXXWORLD

and if I now execute following sequence,

M-x query-replace-regexp RET \(.+\)= RET \,(downcase \1)= RET

the output is,

hello=WORLD
HELLOXXXXWORLD

but if I execute this,

M-x query-replace-regexp RET \(.+\)XXXX RET \,(downcase \1)XXXX RET

then the output is,

HELLO=WORLD
helloXXXXWORLD


> Have a look at (info "(emacs) Replacement and Lax Matches") which
> explains it.  I think the default `case-replace' value of t is the
> problem here.
>

Thanks for pointing me to the right direction. Understood case-replace
thing. But the one thing that is still not clear to me is that, why the
first command doesn't downcase HELLO in the first line, whereas the
second command downcases the HELLO in the second line.

Re: query-replace-regexp on a string with '=' sign

[Tassilo Horn]

Pankaj Jangid <pankaj@codeisgreat.org> writes:

>>> I have a line like this,
>>>   HELLO=WORLD
>>> now I want to downcase the part before the equal (=) sign.
>>> I am approaching like this
>>>
>>>  M-x query-replace-regexp RET \(.+\)= RET \,(downcase \1)= RET
>>>
>>> But this doesn't work.
>>
>> Where "doesn't work" means "doesn't downcase but keeps HELLO", right?
>
> Yes. Suppose I have these lines in a region
> HELLO=WORLD
> HELLOXXXXWORLD
> and if I now execute following sequence,
> M-x query-replace-regexp RET \(.+\)= RET \,(downcase \1)= RET
> the output is,
>
> hello=WORLD
> HELLOXXXXWORLD

For me (with emacs -Q on the current master), it stays as it is, i.e.,
all caps.

> but if I execute this,
> M-x query-replace-regexp RET \(.+\)XXXX RET \,(downcase \1)XXXX RET
> then the output is,
>
> HELLO=WORLD
> helloXXXXWORLD

I can reproduce that.  I think that's documented on the info page I've
cited.  When the replacement is all lower- or all uppercase (like
hello=), `case-replace' has an effect.  But when the replacement is
mixed-case (like helloXXXX), case will be preserved.

>> Have a look at (info "(emacs) Replacement and Lax Matches") which
>> explains it.  I think the default `case-replace' value of t is the
>> problem here.
>
> Thanks for pointing me to the right direction. Understood case-replace
> thing. But the one thing that is still not clear to me is that, why
> the first command doesn't downcase HELLO in the first line, whereas
> the second command downcases the HELLO in the second line.

Huh?  You've said the first command does downcase hello in the first
line although I cannot reproduce that and would not have an explanation
for that behavior.  If it's actually the same for you, then see my last
paragraph for the explanation.

Bye,
Tassilo

Re: query-replace-regexp on a string with '=' sign

[Pankaj Jangid]

Tassilo Horn <tsdh@gnu.org> writes:

> Pankaj Jangid <pankaj@codeisgreat.org> writes:
>
>>>> I have a line like this,
>>>>   HELLO=WORLD
>>>> now I want to downcase the part before the equal (=) sign.
>>>> I am approaching like this
>>>>
>>>>  M-x query-replace-regexp RET \(.+\)= RET \,(downcase \1)= RET
>>>>
>>>> But this doesn't work.
>>>
>>> Where "doesn't work" means "doesn't downcase but keeps HELLO", right?
>>
>> Yes. Suppose I have these lines in a region
>> HELLO=WORLD
>> HELLOXXXXWORLD
>> and if I now execute following sequence,
>> M-x query-replace-regexp RET \(.+\)= RET \,(downcase \1)= RET
>> the output is,
>>
>> hello=WORLD
>> HELLOXXXXWORLD
>
> For me (with emacs -Q on the current master), it stays as it is, i.e.,
> all caps.

Oh this is my mistake. I wanted to point out that only.. It stays caps.

>> but if I execute this,
>> M-x query-replace-regexp RET \(.+\)XXXX RET \,(downcase \1)XXXX RET
>> then the output is,
>>
>> HELLO=WORLD
>> helloXXXXWORLD
>
> I can reproduce that.  I think that's documented on the info page I've
> cited.  When the replacement is all lower- or all uppercase (like
> hello=), `case-replace' has an effect.  But when the replacement is
> mixed-case (like helloXXXX), case will be preserved.
>
>>> Have a look at (info "(emacs) Replacement and Lax Matches") which
>>> explains it.  I think the default `case-replace' value of t is the
>>> problem here.
>>
>> Thanks for pointing me to the right direction. Understood case-replace
>> thing. But the one thing that is still not clear to me is that, why
>> the first command doesn't downcase HELLO in the first line, whereas
>> the second command downcases the HELLO in the second line.
>
> Huh?  You've said the first command does downcase hello in the first
> line although I cannot reproduce that and would not have an explanation
> for that behavior.  If it's actually the same for you, then see my last
> paragraph for the explanation.

Sorry for creating confusion ... again.

Re: query-replace-regexp on a string with '=' sign

[Tassilo Horn]

Pankaj Jangid <pankaj@codeisgreat.org> writes:

>> Huh?  You've said the first command does downcase hello in the first
>> line although I cannot reproduce that and would not have an
>> explanation for that behavior.  If it's actually the same for you,
>> then see my last paragraph for the explanation.
>
> Sorry for creating confusion ... again.

No problem at all.  While the behavior can be explained, I still think
there should be a way to have something like "yes, replace, but don't
obey case-replace", so bug report might still be appropriate.  (As said,
E RET does that but requires 2 keys...)

Bye,
Tassilo

Re: query-replace-regexp on a string with '=' sign

[Pankaj Jangid]

Tassilo Horn <tsdh@gnu.org> writes:

>> Sorry for creating confusion ... again.
>
> No problem at all.  While the behavior can be explained, I still think
> there should be a way to have something like "yes, replace, but don't
> obey case-replace", so bug report might still be appropriate.  (As said,
> E RET does that but requires 2 keys...)

E RET does the job. Yes. But there should be an intuitive way. Because
asking question will also take another keystroke.

Re: query-replace-regexp on a string with '=' sign

[Tassilo Horn]

Pankaj Jangid <pankaj@codeisgreat.org> writes:

>> While the behavior can be explained, I still think there should be a
>> way to have something like "yes, replace, but don't obey
>> case-replace", so bug report might still be appropriate.  (As said, E
>> RET does that but requires 2 keys...)
>
> E RET does the job. Yes. But there should be an intuitive way. Because
> asking question will also take another keystroke.

That's why I'm urging you to report that as a bug. :-)

Bye,
Tassilo

Re: query-replace-regexp on a string with '=' sign

[Pankaj Jangid]

Tassilo Horn <tsdh@gnu.org> writes:

>> E RET does the job. Yes. But there should be an intuitive way. Because
>> asking question will also take another keystroke.
>
> That's why I'm urging you to report that as a bug. :-)

Sure will do.