| Re: cd to nonexistent directory and up again [message #81792 is a reply to message #81708] |
Wed, 17 October 2012 06:05  |
tom.grydeland
Messages: 51
Registered: September 2012
|
Member |
|
|
On Tuesday, October 16, 2012 12:33:19 PM UTC+2, tom.gr...@gmail.com wrote:
> I am using CD to determine whether a given path exists as a directory, and I was expecting the second case to fail just like the first one does, since there is no difference between them in my mind.
Thanks to all those who replied, but I am afraid you're not touching upon my question. I am sorry I have not made my point clearly enough.
1) In a world where directories can be symbolically linked, it is not necessarily true that './bobo/...' is equal to '.', even when 'bobo' exists. Truncating away 'bobo/..' on the assumption that they are the same is arguably incorrect, but it is certainly confusing if it is done in some cases and not in others!
2) In the regular Unix shells, 'ls -ld bobo/..' and 'ls -ld ./bobo/..' both fail, with the error message that the file or directory does not exist.
3) FILE_TEST shows the same strange difference between 'bobo/..' and './bobo/..':
IDL> print, file_test('./bobo/..', /dir)
1
IDL> print, file_test('bobo/..', /dir)
0
It looks like I will have to do what I was hoping to avoid:
1) given a path like 'foo/bar/baz/../quux/..', use file_test(x, /dir) on every sub-path x:
file_test('foo', /dir)
file_test('foo/bar', /dir)
file_test('foo/bar/baz', /dir)
file_test('foo/bar/baz/..', /dir)
file_test('foo/bar/baz/../quux', /dir)
file_test('foo/bar/baz/../quux/..', /dir)
If IDL cannot be relied upon to fail in the next step when one or more of these path components don't exist, then I must test every one of them.
2) If all of these tests pass, then
cd 'foo/bar/baz/../quux/..', curr=old
cd, old, curr=pwd
3) pwd is now the fully-qualified path to the directory indicated by 'foo/bar/baz/../quux/..'
Cheers,
Tom
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
