detecting drives for windows and linux

[BWill]

Hi, I'm writing a file browser, but I'm not sure how I could go about
detecting the drives available on windows and linux systems (preferably
using the standard modules if possible). I guess I could just try to
list root on each letter of the alphabet for windows and see if it
works, but that seems crude; besides, I want to know what kind of drive
each drive/partition is (i.e. is it local and is it a harddrive or
optical drive).

Thanks,

Brian

 

[Tim Golden]

Hi, I'm writing a file browser, but I'm not sure how I could go about
detecting the drives available on windows and linux systems (preferably
using the standard modules if possible). I guess I could just try to
list root on each letter of the alphabet for windows and see if it
works, but that seems crude; besides, I want to know what kind of drive
each drive/partition is (i.e. is it local and is it a harddrive or
optical drive).

I'm not aware of any cross-platform way of doing this.
On Windows you have a few options, but I'd go down
the WMI route; it just makes life easier for doing this
kind of thing. You might want to start by adapting this
example:

http://tgolden.sc.sabren.com/python/wmi_cookbook.html#percentage_free

by removing the restriction to fixed disk (DriveType=3)

TJG

 

[BWill]

I'm not aware of any cross-platform way of doing this.
On Windows you have a few options, but I'd go down
the WMI route; it just makes life easier for doing this
kind of thing. You might want to start by adapting this
example:

http://tgolden.sc.sabren.com/python/wmi_cookbook.html#percentage_free

by removing the restriction to fixed disk (DriveType=3)

TJG

oh, I wasn't expecting a single solution for both platforms, just some
good solutions

thanks

 

[Max]

oh, I wasn't expecting a single solution for both platforms, just some
good solutions

thanks

Are you aware that this idea is somewhat foreign to Linux? (Maybe you
are and want to do it anyway?) Linux puts the whole file system
(including mounted iPods, ISOs and NTFS drives) in one hierarchy.

--Max

 

[Alex Martelli]

Are you aware that this idea is somewhat foreign to Linux? (Maybe you
are and want to do it anyway?) Linux puts the whole file system
(including mounted iPods, ISOs and NTFS drives) in one hierarchy.

Yes, but you may still want to distinguish (because, for example, hard
linking doesn't work across filesystems, and mv is not atomic then).

Running a df command is a good simple way to find out what drives are
mounted to what mountpoints -- the mount command is an alternative, but
its output may be slightly harder to parse than df's.


Alex

 

[Florian Diesch]

Yes, but you may still want to distinguish (because, for example, hard
linking doesn't work across filesystems, and mv is not atomic then).

Why not use os.stat?

Running a df command is a good simple way to find out what drives are
mounted to what mountpoints -- the mount command is an alternative, but
its output may be slightly harder to parse than df's.

Executing df may be expensive if it needs to read some slow file systems.
Reading /etc/mtab is not difficult and much faster.


Florian

 

[Alex Martelli]

Why not use os.stat?

Each os.stat call gives you information about one file (or directory);
it may be simpler and faster to get the information "in bulk" once and
for all.

Executing df may be expensive if it needs to read some slow file systems.

That's what the -n flag is for, if you're worried about that (although I
believe it may not be available on all systems) -- executing mount is
the alternative (just putting up with some parsing difficulties
depending, e.g., on what automounters may be doing).

Reading /etc/mtab is not difficult and much faster

$ cat /etc/mtab
cat: /etc/mtab: No such file or directory

Oops! BSD systems don't have /etc/mtab... so, if you choose to get your
info by reading it, you've just needlessly destroyed your program's
compatibility with a large portion of the Unix-y universe. popen a
mount or df, and information will be easier to extract portably.


Alex

 

[Florian Diesch]

Each os.stat call gives you information about one file (or directory);
it may be simpler and faster to get the information "in bulk" once and
for all.

Depends on what you want to do. Offen you need informations like file size,
permissions or owner in any case.


For things like creating hardlinks it may be the best to just try it and
catch the exception.

That's what the -n flag is for, if you're worried about that (although
I believe it may not be available on all systems) --

GNU df doesn't have it.
Some of the various version of df at Solaris don't have it too. And they
have at least two different output formats.

executing mount
is the alternative (just putting up with some parsing difficulties
depending, e.g., on what automounters may be doing).

I would prefer mount as ir is cheaper than df

$ cat /etc/mtab
cat: /etc/mtab: No such file or directory

Ok, wasn't that clever. On Solaris it's /etc/mntab, I don't have any BSD around

Oops! BSD systems don't have /etc/mtab... so, if you choose to get your
info by reading it, you've just needlessly destroyed your program's
compatibility with a large portion of the Unix-y universe. popen a
mount or df, and information will be easier to extract portably.

On the other hand I know of at least two boxes (no, not mine) where
ordinary users don't have permission to exec mount.





Florian