First off, thanks for your response. On Sat, Oct 10, 2015 at 12:14 PM, Karl Voelker wrote:

...

On Oct 10, 2015, at 11:45 AM, Dan Stromberg wrote:

If I want to read a list of filenames, each terminated with a nul byte, from stdin (kind of like xargs -0), what's the best way to do that in Haskell? Or am I swimming against the current to do anything but newline-termination?

It looks like all the stdin-specific functions are line-oriented, but you can use "Data.ByteString.hGet stdin”.

http://hackage.haskell.org/package/bytestring-0.10.6.0/docs/Data-ByteString....

I'm very much a Haskell newb, but does Data.ByteString.hGet stdin read a fixed (maximum) number of bytes, rather than a nul terminated sequence of bytes?

Google's top hit is:

...

https://downloads.haskell.org/~ghc/7.6.1/docs/html/libraries/bytestring-0.10...

...

...but I don't see anything about nul termination there. I checked the top ~5 hits, and didn't find much.

I don’t think it matters in this case, but you should be aware that Google does not always find the most recent results when it comes to Hackage packages. (See my link above.

I'll keep that in mind. Thanks. -- Dan Stromberg

On Sun, Oct 11, 2015 at 10:43 AM, Brandon Allbery wrote:

On Sun, Oct 11, 2015 at 1:01 AM, Andrew Bernard wrote:

...

On what operating system can filenames contain newlines?

Any unixlike system. You would need to quote the newline from the shell, but it is perfectly valid.

Valid... Ok Perfectly?? Here's a counter view http://www.dwheeler.com/essays/fixing-unix-linux-filenames.html Summary: Funny chars in filenames is a feature close to a bug in *Nix filesystems So my advice for this would be: - If you have better things to do dont bother - if for some reason you do need to bother, respect Postel's law https://en.wikipedia.org/wiki/Robustness_principle and allow it out more reluctantly than in

Andrew Bernard wrote:

Hi Brandon,

It may be valid, but it is just asking for trouble with many tools and utilities and scripts. Funny, I had to check re UNIX. I have been programming UNIX systems for over thirty years and never even imagined a newline in a filename!

Would the OP say why he needs to use newlines in filenames? Something best avoided. I suppose this is not a Haskell matter, but one does have to ask.

The OP must speak for himself, but as far as I am concerned, no reasonable person uses newlines in filenames, ever. However, the possibility is there, and it may happen that someone unreasonable has created a filename with a newline in it. This may become a security issue if, for example, someone creates a file named "/tmp/foo /etc/passwd bar.log" and a careless system person runs a script as root to remove all files named *.log from /tmp/ and subdirectories. find /tmp -name '*.log' -print | xargs rm is an insanely reckless way to do it, which in this case would cause the removal of the password file. Much better: find /tmp -name '*.log' -type f -print0 | xargs -0 rm or even better: find /tmp -name '*.log' -type f -exec rm {} + In summary, if you have to encode a list of filenames in a byte stream, doing it with zero terminated is the correct way to do it. – Harald

Imants Cekusins wrote:

why not base64 encode input text to produce file names:

https://en.wikipedia.org/wiki/Base64#Filenames

An intriguing idea, but sometimes you just have to accept what the upstream program provides. – Harald