3

I occasionally need to extract a single record from a mysqlbackup

To do this, I first extract the single table I want from the backup...

sed -n -e '/CREATE TABLE.*usertext/,/CREATE TABLE/p' 20120930_backup.sql > table.sql

In table.sql, the records are batched using extended inserts (with maybe 100 records per insert before it creates a new line starting with INSERT INTO), so they look like...

INSERT INTO usertext VALUES (1, field2 etc), (2, field2 etc),

INSERT INTO usertext VALUES (101, field2 etc), (102, field2 etc), ...

Im trying to extract record 239560 from this, using...

sed -n -e '/(239560.*/,/)/p' table.sql > record.sql

Ie.. start streaming when it finds 239560, and stop when it hits the closing bracket

But this isnt working as I hoped, it just results in the full insert batch being output.

Please can someone give me some pointers as to where Im going wrong?

Would I be better off using awk for extracting segments of lines, and use sed for extracting lines within a file?

carpii
  • 541
  • 2
  • 4
  • 12

2 Answers2

2

Try something like this which may be tuned to your example but hopefully will get you on the right track

given an input file like this

INSERT INTO usertext VALUES (101, field2 etc), (102, field2 etc), (103,filed2, etc), ...

sed  's/.*\((102,[^)]*)\),.*/\1/' f
(102, field2, etc)

What this does is split your data into 3 sections, remembers the important section using (...) and substitutes this for the whole line.

  • .* everything before the text you want
  • \((102,[^)]*)\) your interesting text
  • ,.* everything else
user9517
  • 115,471
  • 20
  • 215
  • 297
  • \(102.*)\) will capture everything till last block of ()s not just one block containing 102 and related row values. I tried this using GNU sed 4.2.1 and got different match from one in your code block – Hrvoje Špoljar Oct 07 '12 at 21:17
  • @Hrvojepoljar: Fixed - Thanks - I should have added another block of data to test, like I said ... tuned to the example. – user9517 Oct 07 '12 at 21:56
2

Based on Iain's answer I've come up with this sed trick...

$ sed  's/\(INSERT[^(]*\).*\((239560,[^)]*)\),.*/\1\2/' sample.sql | tee 1-record.sql
INSERT INTO usertext VALUES (239560, 2, 3, 4)
$

-- Variation without using sed... You can extract values of some row/insert that match just that one number.

$ cat sample.sql
INSERT INTO usertext VALUES (101, field2 etc), (102, field2 etc), (239560, 2, 3, 4), (5, 8, 9)
INSERT INTO usertext VALUES (101, field2 etc), (102, field2 etc), (1239560, 2, 3, 4), (5, 8, 9)
$ egrep -o '(\(239560\,[^)]+\))' sample.sql        
(239560, 2, 3, 4)
$

rest of insert can be scripted easily if you use -n switch with grep to catch line of match and then just cut everything up to first (

Hrvoje Špoljar
  • 5,245
  • 26
  • 42
  • Many thanks :-) Both your and Iains sed were equally useful, but your egrep seems to be exactly what I was looking for – carpii Nov 16 '12 at 03:35