UNIX for Dummies Questions & Answers

I must have missed something in the question and answers.
Why not use the file utility ?

I have tried the "file" utility. I agree with popo that it falls down when used in a script. The problem is that it doesn't just divide the universe into text and binary files. Rather it outputs stuff like:

Code:

shared object file
lif file
core file
relocatable object file
lint library
compiled Lisp
c program text
s800 shared executable dynamically linked -not stripped

That is just some of the output that I got from a "file *". In many cases it is wrong. But even if were all correct, think what this would do to my script. I would need a case statement with hundreds of entries to collapse this into just "text" and "binary". That would result in a longer, slower script and what's worse, it would no longer be portable. The "file" command on Solaris doesn't know about s800 executables.

In most cases, you don't want a program to guess what language a script is written in. You just want to know if you can safely grep a string out of it or something.

OK, Perderabo, thanks. Now I understand. However, isn't there a simple case (wrapper) switch that can be used to place the output of file into either a ASCII or binary category? Or, as you suggest, file basically flawed and it is better not avoid all together?

In the real olden days, all text files happened to result something like "c program text", "commands text", etc. And so you could check for "*text*" and it would pretty much work. But this subtlety never was documented on the man page. Too many vendors have made too many extentions to "file" for it to be useful today.

On HP-UX, *text*|FrameMaker*|postscript*|shar* would be needed to handle my home directory. I'm not sure if would handle everything. And I have no faith that it would work with the next release of HP-UX. The FrameMaker stuff bothers me. I don't know what FrameMaker is, but I doubt that any of my files would be sensible input to FrameMaker.

Besides, once shells became powerful enough to classify a file with only built-in commands, it made sense to switch anyway. Scripts that do everything without using external programs run much faster.

Thanks, much appreciated.