Batch edit files

man sed, perhaps? :D

Hehe, I do understand why that kind of logic doesn't work on scripts, as I'm a script guy myself. The author of regex is a masochist I guess... He and all who understand and use regex... :confused:

Generically, with sed, you have a command and, optionally, a range for it to work on.

So, you can break it down as "/start/,/end/command" which means, once you start seeing a line with start in it, perform command until you get a line with end in it. The command I used is c, which is:

Code:

---

      c \
      text  Replace  the  selected  lines with text, which has each embedded
              newline preceded by a backslash.

---

Doesn't tell much I know, but I've seen an example somewhere (for extracting multiple lines data from a file, using !d command).

Oh, and as for the -i option:

Code:

---

      -i[suffix], --in-place[=suffix]
              edit files in place (makes backup if extension supplied)

---

Without it, sed will displays the resulting text on standard output, so you can test it first before going "production".

Peace always,
<jdenny>

Thanks a lot !!

I really need to buy that O'Reilly's regex book.. (suggestions for other books welcome !)

Hi

As much as I understand sed (which is not much :( see below),
there is an instrinsic problem here (I think), due to the one-line/multi-line
character, as it also was mentioned by the_JinX:

We have to replace a full multi-line structure with another one.
Both script files replaces such a structure based on the delimiters,
which is dangerous. The question is, how many

Code:

---

<div id="container">

---

are there? To be safe, I have to assume a lot, because it is not
a very specific div-id. Hence, I have to make sure, that the replaced
text is in an one-to-one agreement with what I replace - the delimiters
give an idea, but it is not unique.

Many words ... :D

Quote:

---

(for extracting multiple lines data from a file, using !d command).

---

Unfortunately, I just know the basics of sed (for example I understand
your nice example jdenny). Can you also scan for multiline-patterns (not
by delimiters)?

Cheers.

Both the sed script and my (revised) bash script react to the first instance of the end delimiter after the start delimiter..

I'm not familiar enough with sed nor regular expressions to tell you eabout multiline text editing, I would think that with \n or \r\n you could just enter the whole thing..

<edit type="add">
The sed script actualy changes all instances of <div id="container"> ... </div> in any given file, mine only changes the first instance..
</edit>

Apparently you're right sec_ware... Actually I jumped in just to kid the_Jinx (and to confuse valhallen) :D

What I did is just to show that the whole bash script written by the_Jinx can be re-written in only 4 lines using sed.

The scanning for multiline patterns that I mentioned is using the same method as above (using the /start/,/end/command syntax of sed). Or did I misunderstand you?

the_Jinx, the O'Reilly's book is a must read. But I find this a decent tutorial: http://www.regular-expressions.info

Peace always,
&lt;jdenny&gt;

Hi

Jdenny, you understood me correctly (I think :D ). The delimiter method is
valid, but unfortunately a bit dangerous here (that's why years ago, I
ended up with a c program). Nevertheless, the_JinX's suggestion "\n" and/or
"\r\n" would be worth a try.

Cheers.

Note: I'm not trying to hijack val's thread... If anyone feel I need to start a new thread, please say so.

Ah, sec_ware, you meant without delimiter, I missed that (important) part. Yes, the delimiter method could be dangerous, but it's not, if you know what you're doing. I mean, it's user's responsibility to construct a valid regex to make sure that the search pattern matches what s/he wants to replace.

In this case, I'm assuming that the text between &lt;div id="container"&gt; and &lt;/div&gt; pair is unique. If it were not, then all instances which have the matching pattern would be replaced with the new text. In some cases, perhaps that's exactly what I want. But, if it turns out that the delimiter pair I choose is not unique where it should be, I need to refine my regex.

And yes, multiline regex is sed is supported, although not "straightforward". The regex code would be ugly and I think it's better to use other (better) "tools" such as Perl to do the job. Let me know if you want an example of multiline pattern in sed.

As aside note, sed is famous (notorious?) for writing ugly code. This piece of code

Code:

---

sed '
/^[why]/!d
/[do]$/!d
/^..[real]..$/!d
/^.[code,]/!d
/^..[anyway?]..$/!d
' /usr/dict/words

---

has won a “Worst Abuse of the Shell” prize at a USENIX conference few years back.

Can you guess the expected result? The author said if your copy of /usr/dict/words matches the one on his machine, you should get the following output:
hello
world

Peace always,
&lt;jdenny&gt;

I havn't got the same version (apparently) and it's not even in the same location..

Code:

---

sed '
/^[why]/!d
/[do]$/!d
/^..[real]..$/!d
/^.[code,]/!d
/^..[anyway?]..$/!d
' /usr/share/dict/words
heard
hoard

---

My bad. I don't even have the words file. I just c&p from here without testing it. I think this is the offending line:
/^..[anyway?]..$/!d

Either change !d into d , or , remove the line altogether...

Peace always,
&lt;jdenny&gt;

lol, removing that line does kinda help..

Code:

---

heard
hello
hoard
holed
world

---

using the d instead of !d is even more fun..

Code:

---

hello
holed
world

---