Unreadable UTF characters when they are loaded in MATLAB R2022b, but they could be readable in R2020b

The table data are Chinese characters and saved in MAT-file by using MATLAB R2020b as attachment. In R2022b, MATLAB can load the table variable (db_Indicators), but the non-ASCII part of data in the table are unreadable.

Unzipped the attachment, and then running In MATLAB R2022b as following

load('database.mat','db_Indicators')
double(db_Indicators.UniNum{1})

It produces as

ans =

       65533          49          46          49

In MATLAB R2020b, however,

It could be correctly displayed as

ans =

        8470          49          46          49

Please help to correct the problem. Greatly thanks!

[Walter Roberson]

On my Intel MacOS, R2024b, I get

237 49 46 49

Corresponding to string 'í1.1'

Answer 1

[Bruno Luong]

It sounds like it related to the bug fix they carried out in 2021b, see @Stefanie Schwarz answer here; This breaks the encioding/decofing compatibility before and after this release.

[国强]

Thanks. The clues here hint of the goodness in coding behaviour. I previously drag the variable (db_Indicators) in the workspace window and drop it into the MAT-file. It seems the bad habit to make the data unreadable and being lost by loading in the updated MATLAB R2022b. If using the save command as

save('database.mat','db_Indicators','-append')

the MAT-file would be free of the messy data in later MATLAB.

Maybe, MATHWORK should give some warning of data vulnerability when users save the non-ASCII data in MAT-file by dragging & dropping.