Using with vim substitute command: lines created are not scanned in future steps. Can they be?

dedec0 · 10-21-2021, 01:55 PM

Quote:

Originally Posted by shruggy

Right.

In Vim parlance, that would be

Code:

:g/.\{81,}/s/.\{72}\S\{,8}/&\r\t/g

I left this command waiting, for a few newer posts, because ":g" is something i do not yet know.

Anyway, it does not work. The result is what i already said in #32.

dedec0 · 10-21-2021, 02:05 PM

Quote:

Originally Posted by shruggy

OK. Then how about

Code:

%:s/\v.{10}\S{,4}($)@!/&\r/g

[ typo in the first 3 symbols, which should have
  been ":%s" instead: ]

:%s/\v.{10}\S{,4}($)@!/&\r/g

No indent, and a few other issues. I said indentation is wanted, when i wrote the algorithm.

Result:

Code:

1234567890
 234567890
 234567890
1234567890
 234567890
 234567890
12345678901234
567890 2345678
90
12345678901234
567890 2345678
90
12345678901234
567890 2345678
90
33 66 999
33 66 999
33 66 999

I carefully created example files to make tests easy, but it seems the only person testing commands with them is me. ):

"A hot thread"...

Anyway, although it seems i left a post or another pass by unnoticed, this is just because i am testing them (and frequently correcting them to work) with my test file. This takes some minutes, at times.

shruggy · 10-21-2021, 02:08 PM

Quote:

Originally Posted by dedec0

because ":g" is something i do not yet know.

Zintz tutorial has a good chapter on it. And Vim help, of course.

Quote:

No indent

Then add indent

Code:

:%s/\v.{10}\S{,4}($)@!/&\r /g

dedec0 · 10-21-2021, 02:34 PM

Quote:

Originally Posted by shruggy

Zintz tutorial has a good chapter on it. And Vim help, of course.

Then add indent

Code:

%:s/\v.{10}\S{,4}($)@!/&\r /g

Why your command is "%:s" instead of "%s"? And surprisingly, it works (it does something, i mean... not evaluating the result, for this thread). /-: Of course, the command seeing in vim history is

Code:

:%:s/\v.{10}\S{,4}($)@!/&\r /g

, with one more ":".

dedec0 · 10-21-2021, 02:44 PM

Quote:

Originally Posted by shruggy

Then add indent

Code:

%:s/\v.{10}\S{,4}($)@!/&\r /g

The other issues continue. So, another soso solution.

File:

Code:

1234567890 234567890 234567890
1234567890 234567890 234567890
12345678901234567890 234567890
12345678901234567890 234567890
12345678901234567890 234567890
33 66 999
33 66 999
33 66 999

Command:

Code:

:%:s/\v.{10}\S{,4}($)@!/&\r /g

(note: "\v" here is just to avoid typing several backslahes?)

Result:

Code:

1234567890
  234567890
  234567890
1234567890
  234567890
  234567890
12345678901234
 567890 2345678
 90
12345678901234
 567890 2345678
 90
12345678901234
 567890 2345678
 90
33 66 999
33 66 999
33 66 999

Notice the result lines longer than N (N=10 here). Fine... we just adjust the N to be a bit different. And there is a strange difference in the size of resulting lines, depending on the reason they are splitted.

shruggy · 10-21-2021, 02:54 PM

Quote:

Originally Posted by dedec0

(note: "\v" here is just to avoid typing several backslahes?)

Yep.

Quote:

Originally Posted by dedec0

And there is a strange difference in the size of resulting lines, depending on the reason they are splitted.

That one is easy: don't include the trailing white space

Code:

:%s/\v(.{10}\S{,4})\s*($)@!/\1\r /g

dedec0 · 10-21-2021, 02:59 PM

Quote:

Originally Posted by shruggy

See 4.1 Addresses overview in the GNU sed manual.

Quote:

An address range is specified with two addresses separated by a comma (,). Addresses can be numeric, regular expressions, or a mix of both. The following command replaces the word ‘hello’ with ‘world’ only in lines 4 to 17 (inclusive):

Code:

sed '4,17s/hello/world/' input.txt > output.txt

"Addresses can be numeric, regular expressions, or a mix of both."

So, if i try

Code:

cat input |sed '14,/apple/s/hello/world/'

what does it mean? Is it just "from line 14 to the last line containing 'apple' "? Or will it also count intervals (from lines before 14 that contain "apple") in them, until line 14?

Do you think the documentation can be improved by adding this information?

shruggy · 10-21-2021, 03:01 PM

Quote:

Originally Posted by dedec0

So, if i try

Code:

cat input |sed '14,/apple/s/hello/world/'

what does it mean? Is it just "from line 14 to the last line containing 'apple' "?

It means from line 14 to the first line after it containing 'apple'. After it is important. 'apple' in the line 14 itself won't count

Code:

seq -f'%2.f hello' 15|sed '14s/$/ apple/'|sed 14,/apple/s/hello/world/

shruggy · 10-21-2021, 03:13 PM

Quote:

Originally Posted by dedec0

Why your command is "%:s" instead of "%s"?

Just a typo. Corrected.

dedec0 · 10-21-2021, 03:29 PM

Quote:

Originally Posted by shruggy

That one is easy: don't include the trailing white space

Code:

:%s/\v(.{10}\S{,4})\s*($)@!/\1\r /g

"(anything, including spaces, 10 times; not spaces, 4 at most, as least as possible); zero or more spaces; not EOL" -> substitute with the first group; newline.

This seems to be something we (or i) have done before. But it works for some cases:

File:

Code:

1234567890 234567890 234567890 234567890 234567890 234567890
1234567890 234567890 234567890 234567890 234567890 234567890
12345678901234567890 234567890 234567890 234567890 234567890
12345678901234567890 234567890 234567890 234567890 234567890
12345678901234567890 234567890 234567890 234567890 234567890
33 66 999
33 66 999
33 66 999

Command:

Code:

:%s/\v(.{10}\S{,4})\s*($)@!/\1\r /g

Result:

Code:

1234567890
 234567890 2345
 67890 23456789
 0 234567890
 234567890
1234567890
 234567890 2345
 67890 23456789
 0 234567890
 234567890
12345678901234
 567890 2345678
 90 234567890
 234567890 2345
 67890
12345678901234
 567890 2345678
 90 234567890
 234567890 2345
 67890
12345678901234
 567890 2345678
 90 234567890
 234567890 2345
 67890
33 66 999
33 66 999
33 66 999

But the same command, for this other test file, does something wrong. I do not understand why. File:

Code:

00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Result:

Code:

00 01 02 03
 04 05 06 07
 08 09 10 11
 12 13 14 15
 16 17 18 19
 20 21 22 2
 3
00 01 02 03
 04 05 06 07
 08 09 10 11
 12 13 14 15
 16 17 18 19
 20 21 22 2
 3
00 01 02 03
 04 05 06 07
 08 09 10 11
 12 13 14 15
 16 17 18 19
 20 21 22 2
 3
00 01 02 03
 04 05 06 07
 08 09 10 11
 12 13 14 15
 16 17 18 19
 20 21 22 2
 3

And i discovered that this command can give bad results if there is a space in the end of input lines. So, it is better to remove them, before using.

shruggy · 10-21-2021, 03:56 PM

Quote:

Originally Posted by dedec0

not spaces, 4 at most, as least as possible

as many as possible. This is what greedy means.

Quote:

Originally Posted by dedec0

But the same command, for this other test file, does something wrong. I do not understand why.
...
Result:

Code:

00 01 02 03
 04 05 06 07
 08 09 10 11
 12 13 14 15
 16 17 18 19
 20 21 22 2
 3

Because of the "not EOL" condition. On the last iteration for each line, .{10} will match the ten characters in

Code:

20 21 22 23

\S{,4}($)@! will try to match up to four next characters unless they are followed by EOL. But 3 is followed by EOL, so it's unmatched.

chrism01 · 10-21-2021, 10:56 PM

I have qn; is this just going to be a one time thing(?) in which case i guess a vim cmd will be fine, if you can find one...

If you are going to need to do this > 5 (for example) times, frankly I'd write a short program to do it, which would then be repeatable, easily changed and not prone to fat fingering...
Personally I'd use Perl, but others here would likely use eg awk or or python.

dedec0 · 10-22-2021, 04:41 PM

Quote:

Originally Posted by chrism01

I have qn; is this just going to be a one time thing(?) in which case i guess a vim cmd will be fine, if you can find one...

If you are going to need to do this > 5 (for example) times, frankly I'd write a short program to do it, which would then be repeatable, easily changed and not prone to fat fingering...
Personally I'd use Perl, but others here would likely use eg awk or or python.

No. Not just one time. It will probably have a fair frequency. My most usual plan, to use it, is to format terminal sessions for fora threads or for mailing lists. But understanding what the command does, and how it works, i do not mind rewriting it every time. It will be natural, like using regexes are, today, for me.

pan64 · 10-25-2021, 01:02 AM

I still think it is not a job for sed (because there is no correct search/replace expression for this issue).
That's why I suggested fold or the python textwrap module (or something similar) which can do it perfectly without rewriting, both from command line and from editor.

allend · 10-25-2021, 09:00 AM

There is no _single_ correct search/replace expression for this issue.
It can be done with sed, but requires multiple search/replace expressions.

File:

Code:

1234567890 234567890 234567890
1234567890 234567890 234567890
1234567890 234567890 234567890
12345678901234567890 234567890
12345678901234567890 234567890
12345678901234567890 234567890
33 66 999
33 66 999
33 66 999

Then

Code:

sed -E 's/^([^\t]{10})\s(.{9})\s(.*$)/\1\n\t\2\n\t\3/;s/^([^\t]{14})(.*)\s(.*$)/\1\n\t\2\n\t\3/' input.txt

produces

Code:

1234567890
        234567890
        234567890
1234567890
        234567890
        234567890
1234567890
        234567890
        234567890
12345678901234
        567890
        234567890
12345678901234
        567890
        234567890
12345678901234
        567890
        234567890
33 66 999
33 66 999
33 66 999

Breaking that down a bit:
"s/^(.{10})\s(.{9})\s(.*$)/\1\n\t\2\n\t\3/" targets the first three lines in the file, grouping the non-space characters into three for use as back references.
"s/^([^\t]{14})(.*)\s(.*$)/\1\n\t\2\n\t\3/" targets the second three lines in the file, grouping into three the first 14 characters, then the characters up to 'space' and then remaining characters to end of line for use as back references.
(The third three lines do not require processing and so are not targeted.)

The back references are output with new lines and tab characters to provide indentation as required.

PS - Of course this code breaks for other examples suggested by the OP.