Need to capture part of file name into variable

Question

I have bunch of CSV files named like so in a directory:

 s_m_u_supplemental_fields_account_9_4_2015_11_26_Error.csv
 s_m_u_supplemental_fields_account_9_4_2015_12_3_Error.csv
 s_m_u_academic_status_loa_8_29_2015_0_1_Error.csv

I need to capture the "s_m_u_academic_status_loa" part to append to a column inside the file. I tested the expression "s/^\(s_([^_0-9]+_)+\)\([^\.]+\.csv\)$/\1/" in RegEx Tester online and does capture what I need, but for some reason it fails to work from AIX ksh.

Here is my script so far (AIX 6.2):

#!/bin/ksh

ls *_Error.csv | while read i
do
    echo ${i}
    edfilename = `echo ${i} |sed -e "s/^\(s_([^_0-9]+_)+\)\([^\.]+\.csv\)$/\1/"`
    echo ${edfilename}
done

Answer 1

You might try this version of your script so far

#!/bin/ksh
ls *_Error.csv | while read i
do  echo "${i}"
    edfilename=$(echo "${i}" | sed -e 's/^\(s_([^_0-9]*_)*\)\([^\.]*\.csv\)$/\1/')
    echo "${edfilename}"
done

I dont think aix sed understands + so I've replaced them with *. If you really need + then use \{1,\} which may work on aix.

A simpler pattern that should suffice: 's/\([^0-9]*\)_.*/\1/'