Identifying a text file as UNIX or DOS format

[radiocam]

Hello,
I would be very grateful for some help from the VBA I/O-experienced on this forum.

I'm writing an application to do some conversion on log file timestamps. I have it mostly completed, but am stuck on one issue: These files are text files which could have come from Windows/DOS or UNIX.
I am reading them using sequential access (Open filename for Input as #n, etc)

If they are Windows, it is pretty straightforward:

• text file lines are terminated by CR/LF (or vbCrlf) and Excel reads them fine, splitting up neatly into records

If they are from UNIX, then I have a problem.

• The records are terminated just with an LF which VBA does not recognize as a terminator, So it reads the whole file in as one humongous record.

Now this is okay. In the UNIX streams I can replace the LFs with CR/LFs and write out the file using binary I/O. It works.

Heres my problem: I don't know in advance whether a file is going to be UNIX or DOS format. And what I'm looking for is the most neat/effective way to determine this.
I can do something like count the records, and if there are more than one, conclude that it is a DOS file. But that seems clumsy, and these files are pretty big. The first one I tested was UNIX and gave me one string of length 800,000+ .

Any suggestiobns/solutions?
Cheers and many thanks in advance


For completeness, here's my code for the UnIX to DOS conversion:

Please Login or Register  to view this content.

[snb]

I don't think functions are made for this.
Use a normal macro with arguments instead.

Please Login or Register  to view this content.

[TMS]

What would be the typical length of a record in DOS and what would be the typical length of a record in UNIX? I'm assuming DOS would be relatively short and UNIX would be relatively long. So, read a record, check how long the record is and, if it's greater than a predetermined length, it's UNIX.

Just a thought.

Regards

[radiocam]

Thanks guys.

@snb: A real elegant approach. (I had a function rather than a sub more by accident than intent.)

I have to tweak your replace strings slightly: The unix files just have an LF, the DOS files have a CRLF. With your suggested replacements I'd wind up with just single CRs.

This string would give me CRLFs:

Please Login or Register  to view this content.

Your approach is real neat. I now need to experiment with a few of the larger files to check performance, etc - but thanks. Nice resolution!

Cheers

[snb]

I can't remember the sequence vbcr & vblf or vblf & vbcr

Please Login or Register  to view this content.