Creating a regex to parse a build version

I'm tyring to grab a build verson from a file that contains the following line:

 <Assembly: AssemblyVersion("004.005.0862")> 

and I would like it to return

4.5.862

I'm using sed in dos and got the following to spit out 开发者_开发问答004.005.0862

echo "<Assembly: AssemblyVersion("004.005.0862")>" | sed "s/[^0-9,.]//g"

How do I get rid of the leading zeros for each part of the build number?

The regex to do this in a single step looks like this:

^.*"0*([0-9]+\.)0*([0-9]+\.)0*([0-9]+).*

with sed-specific escaping and as a full expression, it becomes a little longer:

s/^.*"0*\([0-9]\+\.\)0*\([0-9]\+\.\)0*\([0-9]\+\).*/\1\2\3/g

The regex breaks down as

    ^             # start-of-string
    .*"           # anything, up to a double quote
    0*([0-9]+\.)  # any number of zeros, then group 1: at least 1 digit and a dot
    0*([0-9]+\.)  # any number of zeros, then group 2: at least 1 digit and a dot
    0*([0-9]+)    # any number of zeros, then group 3: at least 1 digit
    .*            # anything up to the end of the string

Maybe ... | sed "s/[^0-9]*0*([1-9][0-9,.]*)/\1/g". I'm using a subpattern to filter out the part you need, ignoring leading zeros and non-numeric characters.

There are probably many more clever ways, but one that works (and is reasonably easy to understand) is to pipe it through additional calls:

echo "version(004.005.0862)" | sed "s/[^0-9,.]//g" | sed "s/^0*//g" | sed "s/\.0*/./g"