sed does not replace all the matches

Question

I have this on a file:

Structural Attributes:    text
                          text_id              [A]
                          text_lleng_tr        [A]
                          text_lleng_or        [A]
                          text_cpr             [A]
                          text_for             [A]
                          text_ftr             [A]
                          text_indexador       [A]
                          text_dif             [A]
                          text_reg             [A]
                          text_esp             [A]
                          text_tem             [A]
                          text_tipus           [A]
                          text_data_or         [A]
                          text_data_tr         [A]
                          text_autor           [A]
                          text_traductor       [A]
                          text_titol_or        [A]
                          text_titol_tr        [A]
                          s
                          s_id                 [A]
                          enty
                          contrac
                          contrac_forma        [A]
                          abr
                          date
                          p

Then I run this sed command on it:

sed -i "N;s/\[A\]\n/,/g" file

From which I get:

Structural Attributes:    text
                          text_id              ,                          text_lleng_tr        [A]
                          text_lleng_or        ,                          text_cpr             [A]
                          text_for             ,                          text_ftr             [A]
                          text_indexador       ,                          text_dif             [A]
                          text_reg             ,                          text_esp             [A]
                          text_tem             ,                          text_tipus           [A]
                          text_data_or         ,                          text_data_tr         [A]
                          text_autor           ,                          text_traductor       [A]
                          text_titol_or        ,                          text_titol_tr        [A]
                          s
                          s_id                 [A]
                          enty
                          contrac
                          contrac_forma        ,                          abr
                          date
                          p

So, as you can see there still are [A]+line jump left, if I run the sed command a second time on the same file the amount of [A]+line jump decreases, but I have to run the command 3 times for the [A]+line jump to disappear. So the question is, am I doing something wrong or what is the correct way to do it for the [A]+line jump to be replaced at once.

The N is to act multiline because I actually want to delete the `\n` that comes after the [A]. I have tried the command without the g, but it doesn't help — Andrés Chandía, Nov 24 '17 at 14:20
I can suggest a working awk solution if awk is allowed for you. You could post `awk --v` — RomanPerekhrest, Nov 24 '17 at 14:32
@AndrésChandía, and still posting `awk --v` output would be also desirable — RomanPerekhrest, Nov 24 '17 at 14:34
Yes, of course, then I see how it suits to my needs... thanks again — Andrés Chandía, Nov 24 '17 at 14:38

score 2 · Accepted Answer · answered Nov 24 '17 at 14:35

sed's not a great tool for that task because it's a line-based text edition tool.

You had to use N to consume an additional line so that your working buffer would contain the linefeed separating both lines ; the problem is that this only gives you the ability to work on half the linefeeds each time because once the line is consumed by N it won't be consumed by the next pass of sed : your first pass replaces the linefeed between line1\nline2, then the second pass the one between line3\nline4.

One possible solution with sed would be to consume the whole file in memory before doing the replacement, as shown in this answer :

sed -i ':a;N;$!ba;s/\[A\]\n/, /g' file

That's not a great solution though (especially if your files might get bigger). You might want to wait for another more efficient solution to accept :) — Aaron, Nov 24 '17 at 14:37

sed does not replace all the matches

1 Answers1