Shell Programming and Scripting

Hi,

I have many data files formatted in "dataName-versionNumber-releaseNumber.location.data" in a directory of Linux CentOS 6.2 system.

My directory could have following files for example:

graphicData1-1.2.3-2.0.0.mel.au.data
graphicData1-1.2.3-2.0.1.mel.au.data
graphicData1-1.2.3-2.0.2.mel.au.data

All the files have the same version number 1.2.3, but different release number such as 2.0.0, 2.0.1, 2.0.1. I need to keep the highest release number such as graphicData1-1.2.3-2.0.2.mel.au.data and to delete files with smaller release number, such as graphicData1-1.2.3-2.0.0.mel.au.data and graphicData1-1.2.3-2.0.1.mel.au.data.

How can I do it in a bash script?

Thank you.

Kind regards.

In the format:

Code:

dataName-versionNumber-releaseNumber.location.data

you say that "." is the separator between releaseNumber and location, and it is also the separator between location and data.

I assume that the ".data" is a constant.

Is location always "mel.au" or does it vary between files? If it varies between files, will it be a constant for a given dataName and versionNumber pair, or can you have multiple locations for a given dataName and versionNumber pair?

If there are multiple locations for any dataName and versionNumber pair, and "." is the separator between releaseNumber and location and also appears in both releaseNumber and location, how do we know where releaseNumber ends and location starts? Does any alphabetic character ever appear in a releaseNumber? Does any location ever contain a numeric character? Does releaseNumber always contain two "." characters (as in 2.0.0)? Does location always contain one "." character (as in "mel.au")?

If any of Don Cragun's questions apply, the following will not necessarily work:

Code:

$ ls | sort -r | sort -u -t- -k2,2
graphicData1-1.2.3-2.0.2.mel.au.data

If any of the three numeric components of the releaseNumber could be more than a single digit (e.g., 2.0.10), the sort will have to be much more complex than this. I was also under the impression that there were likely to be multiple dataName values in files in the directory (which would require using -k1,2 instead of -k2,2). If there are multiple dataNames or varying numbers of periods in versionNumbers or releaseNumbers, I'm not even sure that this can be done by sort without some pre-processing by something like awk to generate a consistent sort key at the start of each line.

I think the question is deceptively simple for what may be a much more complex problem. We need more details on the actual filename format.

@Don Cragun: Absolutely! I started fiddling around with awk, thinking about the more-than-one-digit (sub-)release numbers, and it became awkward. What I proposed above is just dealing with the simple sample posted in #1.

The ".data" is a constant. The release numbers only contains numeric characters, no alphabetic characters. It is separated by "." character, but it could be two, or tree or four, such as 2.0, or 2.0.0, or 2.0.0.1.

The location always contain one "." character. The dataName is a single string, the dataName, versionNumber and releaseNumber are delimeted by "-", but after the releaseNumber, the "." is delimeted between the releaseNumber and localtion. The localtion can be treated as constant as well.

Thanks RudiC and Don Cragun.

OK. That helps, but one more point is crucial. Will any part of the versionNumber or releaseNumber ever be more than one digit (such as 2.1.10, 1.11, or 3.0.12.5)?