Hi all


This question is closely related to this one.

Say I got a value of 15347, how do I get this to display as 15.347 by using sed?

I tried:

sed "s|[0-9][0-9][0-9]$|\.[0-9][0-9][0-9]|g"


That, though, gives me this:

15.[0-9][0-9][0-9]


Any ideas?

You did not post any follow up to that thread, so we don't know if you resolved the issue, following the given suggestions.

Anyway, in sed you have to use a pattern reference. That is you have to embed in escaped parentheses the part of the pattern you want to remember, then use \1 as the first pattern, \2 as the second pattern and so on. Your line should be modified as follows:
For more details about this method, see the must-read Sed - An Introduction and Tutorial.

Regarding your original question in the previous thread: to sum floating point numbers without bc you can use awk (as already suggested by H_TeXMeX_H) or even python (by ErV). In awk I would do something more efficient as:
You can try awk also as alternative to the sed command above:
this will give more flexibility about the input number and the output format (you can chose how many digits before and after the dot).

Indeed, and I'm terribly sorry about that.
That's my hard-headedness coming out as I already had a sort of idea how I wanted to tackle this without resorting to another language.

I'm crazy about sed, and I really wanted to be able to tackle my first problem by getting this question answered first. Maybe even learn something more about sed in the process (which I just did by reading your advice).

The plan:

Take out the decimal characters, add the numbers together, then just insert the decimal back in three numbers from the end.


No idea if it's going to work, but here's hoping.


Thanks for your help so far as well, much appreciated.

No need to sorry... really. Furthermore I noticed it was a very recent thread, just after I posted to this one.
Theoretically it should work. Practically it would be very difficult to take in account the different number of digits in each value using sed. Again I'd suggest a more efficient way using awk, since it can do floating point calculations in a straightforward manner.

Hi colucix


Just want to say thanks again for that brilliant piece of awk script you provided. I'm using it extensively at the moment!

Interesting, I didn't know it could be put on one line like that. Still, it's not that readable, where is sum initialized ?

Awk initializes variables to 0, if not given a specific value. To be sure, you can always add the begin section:
but it is not really necessary.

Here is the relevant section from the official AWK programming guide (section 5.3.1)