Detecting 'binary' in a variable

kirbyk has asked for the wisdom of the Perl Monks concerning the following question:

I have an application running under Apache/Mod-perl where a user can upload a csv file. The file gets uploaded, and sits in a perl variable, eventually to be loaded into an Oracle CLOB.

I want to detect if the file they've uploaded is binary or text only, and give the user a helpful error message. (Like, if they upload a .xls file.) Note that the file never exists on a filesystem, so I can't use any unix tricks (and I don't want to write out a temp file.)

I figure I can go character-by-character in a loop and look at the ascii values, but that seems horribly inefficient. Is there a quick regex that could do this check? I'm not worried about Unicode characters, but it'd be nice if extended ascii characters through, say, 165 (to get all the accented characters.)

-- Kirby, WhitePages.com

Comment on Detecting 'binary' in a variable

Replies are listed 'Best First'.
Re: Detecting 'binary' in a variable by Transient (Hermit) on Jul 05, 2005 at 17:49 UTC
would a simple `if ( $file =~ /[^\x00-\xA5]/ ) { # binary } else { #text }` suffice? Update: Also looks like there's a CGI::UploadEasy method "fileinfo" (in case you're using or could use that module)	[reply] [d/l]
Re^2: Detecting 'binary' in a variable by dbwiz (Curate) on Jul 05, 2005 at 18:19 UTC
Not always `$ perl -le '{local$/; $_=<>;}print /^[\x00-\xA5]/ ? "binary" : "text" +' \ /mnt/win/WINDOWS/system32/command.com text` [download]	[reply] [d/l]
Re^3: Detecting 'binary' in a variable by Transient (Hermit) on Jul 05, 2005 at 18:23 UTC
That's correct, but that's not the same regexp: `/^[\x00-\xA5]/` [download] ne `/[^\x00-\xA5]/` [download]	[reply] [d/l] [select]
Re^4: Detecting 'binary' in a variable by dbwiz (Curate) on Jul 05, 2005 at 18:31 UTC
Re^5: Detecting 'binary' in a variable by Transient (Hermit) on Jul 05, 2005 at 18:36 UTC
Re^2: Detecting 'binary' in a variable by kirbyk (Friar) on Jul 05, 2005 at 18:01 UTC
Thanks, that regex does the trick. -- Kirby, WhitePages.com	[reply]
Re: Detecting 'binary' in a variable by brian_d_foy (Abbot) on Jul 05, 2005 at 18:01 UTC
In Perl 5.8, you can open a virtual filehandle on a scalar reference. That might do the trick for you. If you are using an older perl, Tie::Handle::ToMemory does the same thing. You could then use the file test operators, or something like File::Magic. If that doesn't work for you, you can try to match a specific signature for an Excel file (or whatever you might get) with what you see in the uploaded data, but that's a lot more work. Good luck! -- brian d foy <brian@stonehenge.com>	[reply]


go ahead... be a heretic
	PerlMonks