help with extracting entries from a multiple entry columns in a file

DavinaP · 06-30-2017, 04:29 AM

Hi All,
I have a tab separated file where the entries from the 2nd column onwards are multiple and separated by a ;.
I would like to get only the 1st entry of each column. I have 8000+ columns and not giving all here.
Here is a sample:

rs1
AG;0.79780;0.132;0.204;487;923
GG;0.79780;0.115;0.161;213;457
AG;0.79780;0.095;0.152;375;835

I would like to have

rs1
AG
GG
AG

Appreciate any help.
Thank you,
Davina

syg00 · 06-30-2017, 04:39 AM

Trivial in awk - nominate ";" as field separator.

Turbocapitalist · 06-30-2017, 04:40 AM

That should be pretty easy with awk or perl, and there are several ways to approach the problem in either. Which one are you trying and can you show how far you have gotten?

Turbocapitalist · 06-30-2017, 04:43 AM

Quote:

Originally Posted by syg00

Trivial in awk - nominate ";" as field separator.

The sample above has only one column. The other columns are apparently separated by tabs.

So I'd keep tabs as the separator, but use gsub() to zap everything starting with the first semicolon in each field. But is there a way to do that or otherwise get the same result without needing a loop to go through the fields in each row?

DavinaP · 06-30-2017, 04:45 AM

Quote:

Originally Posted by Turbocapitalist

That should be pretty easy with awk or perl, and there are several ways to approach the problem in either. Which one are you trying and can you show how far you have gotten?

Thanks, I have not gotten anywhere much except trying this command:
tr -s '; ' '\t' < "file name".
However that splits each column into multiple columns at the points where ; occurs.
I just want the first entries of each column (remember I have thousands of columns).

syg00 · 06-30-2017, 04:51 AM

Oops - didn't read that too well did I. Sorry about that. I'll be back.

Turbocapitalist · 06-30-2017, 04:52 AM

Ok. Try escalating to awk then. Be sure to see the manual page.

Code:

man awk

But that is a reference (actually the reference) only and though you should use it a lot, it might not be the best place to start with awk. So also see this site:

http://www.grymoire.com/Unix/Awk.html

It is a very thorough introduction.

syg00 · 06-30-2017, 05:45 AM

Quote:

Originally Posted by Turbocapitalist

So I'd keep tabs as the separator, but use gsub() to zap everything starting with the first semicolon in each field. But is there a way to do that or otherwise get the same result without needing a loop to go through the fields in each row?

gensub maybe - that way you can use back-references.
Personally I'd use sed - same/similar regex.

syg00 · 06-30-2017, 06:31 AM

How's your regex fu Davina ?.
Your data (for this discussion) can be defined as "a bunch of non-semicolon characters (that you want to keep), followed by a bunch of non-whitespace characters (that you want to remove)". Define that in regex, and make the substitution global.

Turbocapitalist · 07-01-2017, 12:41 AM

Quote:

Originally Posted by syg00

Personally I'd use sed - same/similar regex.

Yes. If one thinks about the lines as a single unit, then sed is a good idea. I had been thinking about the line as a record with fields and thus gravitated to awk. Either will work. The language sed is a little terse while awk is a little more complicated, though.

DavinaP, the substitution command in sed is what to look at:

Code:

sed -e 's/old/new/g;' < oldfile.txt > newfile.txt

The greater than > and less than < signs are IO redirects in the shell.

BW-userx · 07-01-2017, 07:40 AM

I do not use awk but try this,

Code:

userx%slackwhere ⚡ testDIR ⚡> awk -F\;  '{print $1}' fileDirLit
rs1
AG
GG
AG

http://cs.canisius.edu/ONLINESTUFF/P...K/awk.examples

or to keep it handy

Code:

userx%slackwhere ⚡ testDIR ⚡> awk -F\;  '{print $1}' fileDirLit > results       
userx%slackwhere ⚡ testDIR ⚡> cat results
rs1
AG
GG
AG

to skip that first line

Code:

userx%slackwhere ⚡ testDIR ⚡> awk -F\; 'NR > 1 {print $1}' fileDirLit 
AG
GG
AG
GG
AG

AwesomeMachine · 07-01-2017, 01:02 PM

You can use:

Code:

$ tr ';' \t < file > file2
cat file2 | awk '{print $1}' > file3

That is untested, but I think it will work. What you're doing is changing the semicolons to tabs, which are white space, and then selecting the column before the first white space.

allend · 07-03-2017, 08:05 AM

My feeling is that the example given has been confusing.
If the data format is tab separated columns with semicolon delimiters within columns, such as

Code:

AG;0.79780;0.132;0.204;487;923	AG;0.79780;0.132;0.204;487;923	AG;0.79780;0.132;0.204;487;923
GG;0.79780;0.115;0.161;213;457	GG;0.79780;0.115;0.161;213;457	GG;0.79780;0.115;0.161;213;457
AG;0.79780;0.095;0.152;375;835	AG;0.79780;0.095;0.152;375;835	AG;0.79780;0.095;0.152;375;835

then I suggest using awk

Code:

 awk -F ";[^\t]+" '{for (i=1;i<NF;i++){printf"%s", $i}; printf"\n"}' <inputfile>

BW-userx · 07-03-2017, 08:11 AM

in OP post he says all he wants is the very first column which is all of the AG GG AG etc..
which this actually gives him

Code:

awk -F\;  '{print $1}' fileToLooKAt > results

AwesomeMachine · 07-03-2017, 10:32 AM

But the OP doesn't show any tabs, so $1 is the whole row.

Sorry, didn't see the "-F".