Copyfile doesn't preserve timestamp

[status, msg, msgID] = copyfile(filePath1, filePath2);

This line copies a file from a local directory into a USB drive, but the timestamp is not preserved, and the current time is used instead.

Why does this happen if the documentation says the opposite?

MATLAB R2019b, macOS High Sierra 10.13.6

[dpb]

Maybe OS problem? Don't have Mac but symptom doesn't appear on R2019b on Win7 which would seem to imply it's not inside the TMW-specific code. It's builtin so can't see what they did, but one presume it's just passing the arguments to the OS, anyways...

[Image Analyst]

Windows has 3 dates: modified date, created date, and accessed date. Maybe macOS is the same. Have you looked at all 3 dates?

[Stephen23]

Vincenzo'S "Answer" moved here:

I also observed that this behaviour is somewhat file-dependent.

On one file I get the same modified date as the original copy, on other ones (a .dmg file, a .png file) I get the current time as the modified date.

Strangely, the creation date is set for both at an arbitrary date, to which I see no link with the original copies.

[Stephen23]

Vincenzo's "Answer" moved here:

https://it.mathworks.com/matlabcentral/answers/567396-copyfile-doesn-t-preserve-timestamp#comment_942210

Let me disagree here. The documentation doesn't specify an OS-dependent compatibility, so if it's an OS-dependent issue, it means it's an OS-dependent bug in TMW-specific code. It would be useful to compare results to analyze different behaviours and possibly raise a bug fix request to The Mathworks.

Answer 1

[Image Analyst]

Windows has 3 dates: modified date, created date, and accessed date. Maybe macOS is the same. For Windows 10, the modified date is the same as it was. The Created date and the Accessed date of the new file are the current time that copyfile() was executed. The documentation only says "After a successful copyfile operation, the timestamp for the new file is the same as the timestamp for source." I agree this is ambiguous when the operating system maintains 3 dates for the files. It could be more explicit on what "timestamp" means especially when only one of the timestamps is the same and two will be different. You could click the single star at the bottom of the documentation and then put in your suggestion (like I just did). I know people do read those suggestions.

MATLAB also does not seem to have a touch function so perhaps you can use system() to run such a command after you run copyfile(), if your operating system has such a touch function.

[Vincenzo]

Yes, as a workaround I ended up doing exactly that, a system() call to the macOS terminal command

touch -mf YYYYMMDDHHmm filePath

after copyfile() has been executed.

[Image Analyst]

And does that change all 3 timestamps, or only 1 of them? Since you do it, I assume it's changing all 3 but I just want to be clear on that.

[Vincenzo]

From what I see, the touch command on macOS modifies only the "last modified" timestamp.

[Image Analyst]

Which timestamp(s) do you need to be kept the same, and why? What's wrong with the way it's doing it now? What app is going to be inspecting those timestamps? What is it expecting to see?

[Vincenzo]

I'm building a folder synchronization app, and I want a synchronized file to keep its last modified timestamp.

Therefore what I need is exactly what the "touch" terminal command does. The workaround of using the system() call it's ok from a functional point of view, it's nonetheless confusing that the MATLAB function copyfile() does not handle this, and even more strangely that it handles the last modified timestamp only for some files and not for others.

I think there is definitely something to inspect in the behaviour of the copyfile() function on macOS. Alternatively, it has to be clarified in the documentation that the compatibility on macOS is limited.

[dpb]

If you can create a reproducible test case that illustrates a problem, submit a formal bug report to TMW. Answers is NOT official support. See "Contact Us" (the old-style telephone receiver) icon at top of page or email [e-mail]

[Walter Roberson]

For folder synchronization you should consider system() of rsync