Regular expressions (substituting floating point numbers)

[bluesmodular]

Using printf and ed, or some other way, how would I replace a line that contains only a floating point number, e.g. 6.00, with another floating point number stored in variable called $bulksize or $half ?

I would like to identify the line by the line number.

A related example:
In this snippet:

bulksize=`expr 6.00`
half=$(echo "scale=2; $bulksize / 2" | bc)
for i in */run_example; do
	printf '187s/= [^,]*/= %s/\nwq\n' "$bulksize" | ed -s "$i"
	half=$(echo "scale=2; $half + 0.02" | bc)
	bulksize=$(echo "scale=2; $half * 2" | bc)
done

The line

printf '187s/= [^,]*/= %s/\nwq\n' "$bulksize" | ed -s "$i"

works for replacing the number in this line:

  celldm(1)   = 6.00,

Thanks for your help!

[nmset]

To me the problem is not clear enough as laid out. I understand you want to replace floating point values in a file, a sample of the target file would help. printf and ed may not be mandatory. , sed may well do the job.

[Cabhan]

Well, I have to give you credit for using ed. I don't think I've actually ever seen someone use it before.

This will be easier to do by using sed, depending on how exactly we want to do this.

If there's just a single floating point number on a line and we want to replace it, I would personally prefer to use sed. The sed command would be:

sed -re '187 s/[[:digit:]]\.[[:digit:]]/'"$bulksize"'/' "$i"

There's a lot of weird quoting here, but it's necessary.

Here, [[:digit:]]\.[[:digit:]] matches a floating point number, and this replaces it with bulksize on line 187.

Now, if want to replace all of line 187 (maybe it's a lot more complicated, with multiple floating point numbers, or you just always want to change it to a certain line), you could do:

sed -e '187 c NEW_TEXT'

People tend not to use ed these days because it's very complicated and weird. sed tends to be a bit cleaner.