UNIX for Dummies Questions & Answers

Hello,

I have to turn:

Code:

Apple Inc.:325,64:329,57

into

Code:

Apple Inc.:325,64:329,57:3,93

3,93=329,57-325,64.

My code:

Code:

cat beurs.txt | sed 's/\([^:]*\):\([0-9]*\),*\([0-9]*\):\([0-9]*\),\([0-9]*\)/\4\.\5-\2\.\3/' beurs.txt | bc| tr '.' ',' | sed 's/^-*,/0,/' > winstmarges.txt; paste -d: beurs.txt winstmarges.txt; rm winstmarges.txt

works but can I do it without creating temporary files?
Can I use bc to do calculations on a specific field?
Or can I do calculations within sed?

Thanks

Try this...

Code:

awk -F: '{x=$0;gsub(",",".");y=$3-$2;gsub("\\.",",",y); print x OFS y }' OFS=: infile

--ahamed

If you want (slash need) to stick to sed and bc, try this:

Code:

sed 's/^\(.*:\)\([0-9]*\),*\([0-9]*\):\([0-9]*\),\([0-9]*\)/print \"\1\2,\3:\4,\5:\",\4\.\5-\2\.\3,\"\n\"/' beurs.txt|bc|sed 's/\(\.\([0-9]*\)\)$/,\2/'

Explanation: Don't abuse a cat! First turn the line into a string by enclosing it in '"'. Add a comma, the expression, a comma and a newline. Feed this to the print function of bc.

Finally translate the last '.' into a ',', because bc is not aware of locale-issues. In the Netherlands we use a decimal ',' instead of a '.'.

When nesting groups in sed, the deeper nested groups get higher numbers. e.g.: (a(b(c))) \1=abc, \2=bc, \3=c. This way we can get rid of the '.'. Of course this applies not only to sed but to other regular expression environments as well.

I've tested this on Ubuntu 11.10.
Personally I think the above solution is very unreadable. A fairly simple way to achieve your goal with awk could be:

Code:

awk -F: '{printf $0":";gsub(",",".");wm=$3-$2;sub("[.]",",",wm);print wm}' beurs.txt

If this doesn't work in your situation try nawk instead of awk.

Groetjes/Greetings

Eric

Hello,
Thank you!
Yes, for now, I'd like to stick with sed.
So when you have a line with a string in it to calculate (example: test,5.5+5.5)
than you have to enclose the string into "" followed by a comma, the expression you want to calculate, another comma and a newline? (print "test",5.5+5.5,"\n")

I'm from Belgium, so we also use a decimal ',' instead of a '.'.

Greets!
PS: Happy New year!

---------- Post updated at 04:58 PM ---------- Previous update was at 12:00 PM ----------

And what if you have more than one string?
Example:

Code:

test:1+3:unix:1+5

to

Code:

test:4:unix:6

To be honest, I've hardly ever worked with bc before. I just took a glimpse at the man page and came up with this idea. Since we are mixing strings and expressions, the print command seemed the way to go. It prints strings while other things in the argument list are interpreted as expressions before they get printed.

Your latest exercise could be solved as follows:

Code:

echo 'test:1+3:unix:1+5'|sed 's/^\(.*:\)\(.*\)\(:.*:\)\(.*\)$/print \"\1\",\2\,\"\3\",\4,\"\n\"/'|bc

The result is:

Code:

test:4:unix:6

Groetjes/greetings,
Eric

Thank you so much!
I'll give it a try tomorrow.