Fully resolving path names

286 views (last 30 days)
Ray
Ray on 12 Dec 2012
Commented: Alec Jacobson on 8 Aug 2023
Objective: retrieve and store an absolute (or fully-qualified) path name for a file or directory from a user specified string which could be given as absolute, relative to a MATLAB path, relative to the current working directory ('.' or '..'), or relative to a user's directory (~ or ~user).
Builtin functions like EXIST and WHICH do a great job of searching out files that are on the path, hidden in class folders, specified as relative to the current directory, ... but they do not robustly return the full path for generic files (i.e. non-MATLAB user files that may not be on the MATLAB path). WHICH has a somewhat nonsensible behavior of claiming a file that EXIST, DIR, and FOPEN can find does not exist.
The goal is to be able to robustly obtain the absolute path name for a file so that it can be found again even from a different working directory or with a different MATLAB path.
For now my solution is to confirm existence of the file (EXIST) then save both the specified filename and the current working directory. Assuming that the MATLAB path doesn't change, I can cd to the directory, execute FOPEN, and then cd back whenever I need to access the file. Neither clean nor efficient.
Ideally I would like a function absolutePath = resolvePath(filename) that uses the same search criteria as the builtin functions.
Thoughts?

Answers (6)

Ray
Ray on 12 Dec 2012
Not a complete solution for all the smart places the builtin's look, but a useful snippet that should work for most user file and directory applications:
function fullpath = resolvePath(filename)
file=java.io.File(filename);
if file.isAbsolute()
fullpath = filename;
else
fullpath = char(file.getCanonicalPath());
end
if file.exists()
return
else
error('resolvePath:CannotResolve', 'Does not exist or failed to resolve absolute path for %s.', filename);
end
  1 Comment
Alec Jacobson
Alec Jacobson on 8 Aug 2023
This doesn't seem to work correctly for paths starting with ../ on mac. It always thinks the current directory is ~/Documents

Sign in to comment.


Jan
Jan on 12 Dec 2012

Jurgen vL
Jurgen vL on 18 Jun 2020
Using the python os submodule to get an absolute path:
string(py.os.path.realpath(py.os.path.expanduser(<PATH>)))
The above does not yet check if the absolute path exists. Requires python to be installed and discoverable by MATLAB.
  1 Comment
Alec Jacobson
Alec Jacobson on 8 Aug 2023
First call is slow (launching python?) but subsequent calls are fast and this works for paths starting with / , ~/ and ../

Sign in to comment.


Andrea Barletta
Andrea Barletta on 6 Nov 2020
I had a similar question and I realized it was easier than I thought. I am sharing my code, just in case (no Java or Python used). I should work on all versions >=R2017b.
function absPath = getAbsPath(obj)
getAbsFolderPath = @(y) string(unique(arrayfun(@(x) x.folder, dir(y), 'UniformOutput', false)));
getAbsFilePath = @(y) string(arrayfun(@(x) fullfile(x.folder, x.name), dir(y), 'UniformOutput', false));
if isfolder(obj)
absPath = getAbsFolderPath(obj);
elseif isfile(obj)
absPath = getAbsFilePath(obj);
else
error('The specified object does not exist.');
end
  6 Comments
Andrea Barletta
Andrea Barletta on 10 Nov 2020
Unfortunately I am unable to test my function on any version older than R2018a, for which I can confirm it works. I have some (unconfirmed) confidence it also works in version R2017b as I wrote in my first post. It will certainly not work on previous versions as some dependencies would be missing (e.g. isfolder).
Walter Roberson
Walter Roberson on 10 Nov 2020
However, the remaining part of the code needs to be there to address files other than folders, as per original request. Furthermore, dot notation is to be preferred to getfield according to the official documentation:
Yes, dot notation is preferred. However, Tom Hawkin's version using getfield() can be implemented in a single call without making any assignments, and assignments used to be strictly necessary in order to use dot notation without using getfield() or subsref().
Starting in R2019b, you can use dot notation to index the result of a function call, such as
dir('.').folder
However, dir() of a folder returns entries for the folder and its parents and everything directly inside of the folder, so dir('.').folder would be certain to return an expansion into a minimum of two (identical) entries. If you were passing the result into something, you would probably get complaints about too many inputs. For example
>> length(dir('.').folder)
Error using length
Too many input arguments.
getfield(dir('.'), 'folder') would return only a single copy, but using dot notation as "preferred" would have to deal with the multiple copies, such as
struct('folder', unique({dir('.').folder})).folder
But more likely
First = @(varargin) varargin{1};
First(dir('.').folder)
Though of course you could also do
FirstFolder = @(dinfo) dinfo(1).folder;
FirstFolder(dir('.'))

Sign in to comment.


Walter Roberson
Walter Roberson on 12 Dec 2012
Pull off the file-name part. cd() to what remains. pwd() to find out the directory name. cd() back.
  1 Comment
Ray
Ray on 12 Dec 2012
Thanks for the input. Your suggestion does solve some of the issues. However, if the input string is relative to the current working directory, one of the MATLAB paths, or one of the other nooks and crannies that EXIST and other builtins check, your solution will fail because there is no where to CD.

Sign in to comment.


Gavriel Aminov
Gavriel Aminov on 18 Jan 2023
Please examine my File Exchange contribution: Convert local to remote (global) filename.

Categories

Find more on Environment and Settings in Help Center and File Exchange

Products

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!