Matlab's read.m produces an error in version 2025a

I am using a library that has worked seamlessly on previous Matlab versions, up to and including 2024b

At some point, this library calls C:\Program Files\MATLAB\R2025a\toolbox\matlab\graphics\graphics\+matlab\+graphics\+internal\+figfile\@FigFile\read.m

In this function, a variable verString (1x1 string) is created on line 85, using regexp(). Then, read.m attempts to convert it to double using str2double() on line 87. However, str2double will not accept a string array as input. (Inputing foo = "string"; str2double(foo); on the Matlab2025a command line produces an identical error.)

There is no possibility to rewrite this function (it is read-only), so any help with what to do would be greatly appreciated.

[Stephen23]

"(Inputing foo = "string"; str2double(foo); on the Matlab2025a command line produces an identical error.)"

No error here:

foo = "string";
str2double(foo)

ans = NaN

Most likely you have some 3rd-party toolbox function which is shadowing a MATLAB function. Which toolbox and which function might be tricky to figure out. But for a start, look for any that redefine STR2DOUBLE, and remove from the path.

If you require more help then please show us the entire error message. This means all of the red text.

[Jeffrey]

Was found and fixed, thank you. Not sure why/how this never seemed to create a similar error in 2024b, but I'll take the win!

[dpb]

So, enquring minds and all that <vbg>...what, specifically, was found and fixed?

Besides the simple curiosity, it could be helpful if is a 3rd-party issue and someone else is using the same toolbox.

If that were the issue, it would seem a bug report submitted to the vendor would be in order with the "you ought not to have done that, doncha' see!" admonition.

<PS> Off topic, but will note...this is the first time since the forum outage I've been able to respond; until just this morning after the <upstream error> cleared, could view but any attempt to reply either Answer or Comment simply hung...no klew as to why as it clearly was functional for others for several days.

[Image Analyst]

@dpb maybe you needed to clear the browser cache? Sometimes it automatically refreshes pages rather than using the cache. Or maybe "it's just you!" 😆

[Jeffrey]

@dpb what was found was the function shadowing str2double. Indeed in the 3rd party library, there was a function written with that name. I was surprised, though, that having used this toolbox for probably 4 or 5 years that it never caused any incident until the latest version (2025a). In the end, I did not write the library, so I cannot comment on exactly why, but it needs this to run properly; temporarily deleting the shadow function creates a different error. So, in the end I just modified the shadowing function to take "string" as input by adding a case to the switch it uses when evaluating how to deal with different types of input.

[dpb]

That's ugly...and I still suggest the vendor should receive a bug report from an irate user that it is not kosher to alias builtin MATLAB functions. They'll probably not change their practices, unfortunately, but at least they may make their function more tolerant.

I suspect what has happened is that somewhere in its bowels MATLAB changed one of its internal behaviors with one of the string-handling routines to use the new strings class instead of cellstr and that would then propogate on forward with a silent conversion until, as you found, a not-ready-for-strings routine was called. I can't think of the precise syntax right now, but I have been surprised a couple of times that one of my functions began returning a string instead of cellstr after its introduction without me knowingly introducing them.

[dpb]

@Image Analyst, I've no idea. I observed that refreshing the page would bring new content, but still if attempted to edit, comment, answer, the browser would simply hang interminably with the spinning circle...this occurred with both Firefox the default I use and MS Edge.

[Paul]

Should vendors get in the habit of putting all (most? some?) of their functions in a namespace and calling them as such to avoid the shadowing problem?

[dpb]

Since that is now an option, probably, at least for new code. There is the issue of backwards compatibility, though, if supporting older releases.

Forth had variable and function (aka "words") dictionaries "way back when" which is where I first met the concept...most handy, indeed.