Regex help!

[lpanokarren]

Hello monks,

I am having trouble with extracting text from unformatted descriptions. Basically, I need to extract disk capacity from the data center description which basically has a numeric value preceding the unit of measure like GB

E.g. x='Disk - HD 500GB 2.5 SATA (00X3Y)'

I wrote a sed script

echo $x | sed "s/^[^0-9]*\([0-9]*[.]*[0-9]*\)[^0-9A-Z]*GB.*/\1/g"

this works in this case. However, if I run into - x='USE 28-000009 FLASH CARD, IODUO, 1280GB, MLC, FUSION IO, FS3-204-641-CS-0001'

echo $x | sed "s/^[^0-9]*\([0-9]*[.]*[0-9]*\)[^0-9A-Z]*GB.*/\1/g"

returns nothing

Any ideas?

[LanX]

this is Perl

  DB<110>  @a= ('USE 28-000009 FLASH CARD, IODUO, 1280GB, MLC, FUSION 
+IO, FS3-204-641-CS-0001' , 'Disk - HD 500GB 2.5 SATA (00X3Y)', 'Disk 
+- HD 500.31415926GB 2.5 SATA (00X3Y)')
 => (
  "USE 28-000009 FLASH CARD, IODUO, 1280GB, MLC, FUSION IO, FS3-204-64
+1-CS-0001",
  "Disk - HD 500GB 2.5 SATA (00X3Y)",
  "Disk - HD 500.31415926GB 2.5 SATA (00X3Y)",
)

  DB<111> print / (\d+(?:\.\d+)?)GB\W/,"\n" for @a
1280
500
500.31415926

and sed ain't no PCRE!

[lpanokarren]

Thanks a ton Rolf! although, do you know if your code works for this input?

'Disk - HD 600GB 15K'

I used the regex you provided, and it seems for this the output is all 0s - unless I am doing something wrong

[LanX]

> unless I am doing something wrong

you are doing something wrong!

Read more... (17 Bytes)

[Anonymous Monk]

I used the regex you provided, and it seems for this the output is all 0s - unless I am doing something wrong

But you don't show what you're doing

Read more... (3 kB)

[LanX]

> Any ideas?

use <code> tags and ask sedmonks ? =)

edit

but as far as I can decipher the stuff you posted your regex checks for strings with non-numbers from start till "size GB". (this /^[^0-9]*...)

But the failing string has numbers in between.

If you provide Perl code and sample data people here might help you. =)

[shmem]

Basically, I need to extract disk capacity from the data center description which basically has a numeric value preceding the unit of measure like GB
...

echo $x | sed "s/^[^0-9]*\([0-9]*[.]*[0-9]*\)[^0-9A-Z]*GB.*/\1/g"

Why so complicated? This should do:

echo $x | sed "s/.\+ \([0-9]\+\)GB.*/\1/g"

But as LanX says, sed(1) doesn't use PCRE. Read perlrun and perlre to understand the following:

echo $x | perl -ple 's/.+?(\d+)\s?GB.*/$1/'

[lpanokarren]

sorry - for the poor formatting - I am actually trying to implement this in Informatica which accepts Perl compatible regular expressions (PCRE) - so this is technically a problem now I am not able to solve even with PCRE I used sed to illustrate the PCRE I was using - not good at one liners.