Using perl to grep a list of patterns from an input file

Shell Programming and Scripting
1

I have been struggling to grep a file of NGrams (basically clusters of consonants or Consonant and Vowel) acting as a pattern file from an Input file which contains a long list of words, one word per line. The script would do two things:
Firstly read a text pattern from a large file of such patterns: they are all consonant clusters and grep them from the input file which will have one word per line. It would be great if the script could also identify the clusters whether they occur in the beginning, middle or end of the file. But that would be the icing on the cake.
Secondly the output should be sorted on the clusters found. In case a given cluster is not found, it whould be marked as such.
An example would help:
The pattern file is

cr
pl
sl
st
pn

The input file would be

please
crawl
creep
slip
slide
apnea
pneumatic

The desired out put would look like:

#cr
crawl
creep
#pl
please
#sl
slip
slide
#st NONE
#pn
apnea
pneumatic

Is it possible to write a Perl code to do something of the sort?
I have used grep and egrep with the tag to grep from a pattern file, but the data is so huge that the utilities do not give satisfactory results.
Many thanks in advance

2

How big are the files?

3

If grep is running out of memory, you may need to rethink your strategy. Just how large is the pattern file?

Are duplicate results allowed? Are things allowed to appear in more than one list?

If the pattern file's gigantic, you might be in trouble. Otherwise I might try something like this:

awk 'NR==FNR { A[++L]=$1; print "#" $1 > $1; next }
        { for(N=1; N<=L; N++) if(index($0, A[N])) { print > A[N] ; T[N]++} }
        END {
                for(N=1; N<=L; N++)
                if(!T[N]) { close(A[N]); print "#"A[N]" NONE" >A[N] }
        }' listfile inputdata

This will create output files named cr, pl, etc which you can cat together later if you need to.

1 Like
4

The pattern file will have around 300 patterns. It is the Input file i.e. the file which has to be grepped which is the worrisome part. It will have around 300,000 unique records. The grep from pattern gave really goofed up results and this is why I posted the request.
Many thanks for your interest

---------- Post updated at 11:22 PM ---------- Previous update was at 08:29 PM ----------

I tested the data on the large file and it worked perfectly, all the more so since the data is in UTF-8 format.
How can I modify the code if I want it to give me only two to three samples from the input file.
I am new to AWK and I tried to change the

N++

operator but could not make it work for a numeric value.
Many thanks again for your kind help

5

300,000 records is not that large unless they are really, really huge records. And 'really goofed up results' is not the same as 'the data is so huge that the utilities do not give satisfactory results' -- if the data were really too large, it would have said 'out of memory' or some such.

T[N] is how many times its been printed. I've been using it to tell whether I should put NONE in a file. Now I also use it to check if a pattern has been printed enough times to stop printing it any more.

awk 'NR==FNR { A[++L]=$1; print "#" $1 > $1; next }
        { for(N=1; N<=L; N++) if(index($0, A[N])) { T[N]++ ; if(T[N] <= 3) print > A[N] ; } }
        END {
                for(N=1; N<=L; N++)
                if(!T[N]) { close(A[N]); print "#"A[N]" NONE" >A[N] }
        }' listfile inputdata
6

Many thanks for the explanation.
The size of the pattern file has increased but awk seems to respond pretty well. To speed up the process, I converted all Unicode to 8 bit and got a fair amount of speed in computation.
Thanks a lot.

---------- Post updated at 11:45 PM ---------- Previous update was at 11:13 AM ----------

Sorry to hassle again. I got the logic, but I have two queries:
A simple one:suppose I wanted to print all the data found on one line instead of on separate lines, what would I have to tweak ? I tried tweaking the print command but could not get the script to print all words on one line.
More complex. My pattern file has gone huge. I am testing 4 letter combos which means around 64000 strings to match. When I run the file against around a similar size of inputs, the script executes but does not give the expected result.
Any solutions. I am quite desperate to solve this issue. Many thanks in advance but with my limited knowledge of AWK and the reading I am doing from OReilley's book on sed and awk, things do not seem to work out.