Writing unicode files

[MBaas]

I just noticed a problem when I tried to process that I exported from XY. For a repro, try this:

Code: Select all

    writefile("foo1.txt","""Mein Behälter"" C:\Data\Mein Behälter\",,"utf8bom");

Open foo1.txt with a binary editor and examine the two "ä"s: the first one is written as "C3 A4" - the second one as "61 CC 88".

I suppose this is "somewhat legal" - but it's not entirely correct:
VS Code (which is very tolerant) has no problems with it, UltraEdit can also handle it - but its hex mode shows there's some weirdness.

05-05-2024_15-02-01.png

And the environment I want to read it from simple injects blanks - which messes up subsequent processing, ofc.

Unfortunately I don't know this stuff well enough to provide a better explanation - but I hope it is reproduceable and fixable anyway

[Horst]

You write UTF 8 with BOM.
So get the BOM characters, or what else do you expect ?

[admin]

Looks like a bug, but I ran your script and the resulting file looks as it should:

2024-05-05_164411.png

I'm using the English locale though, which might (but shouldn't!) make a difference:

Code: Select all

System / Thread Locale ID: 1033 (en-US) / 1033 (en-US)
Default ANSI Code Page: 1252  (ANSI - Latin I)
Active ANSI Code Page: 1252  (ANSI - Latin I)
Default OEM code page: 437   (OEM - United States)
Active OEM Code Page: 437   (OEM - United States)

[MBaas]

Just to confirm...I changed my system settings - but it did not affect the result of writefile!

My settings now match yours: (at least we know now that THIS did not cause the difference - as you said!)

Code: Select all

System / Thread Locale ID: 1033 (en-US) / 1033 (en-US)
Default ANSI Code Page: 1252  (ANSI - Latin I)
Active ANSI Code Page: 1252  (ANSI - Latin I)
Default OEM code page: 437   (OEM - United States)
Active OEM Code Page: 437   (OEM - United States)

[admin]

Try this:

Code: Select all

text utf8encode("""Mein Behälter"" C:\Data\Mein Behälter\", 1, 1);

[MBaas]

Code: Select all

"Mein Behälter" C:\Data\Mein Behälter\

[admin]

And this:

Code: Select all

text hexdump(utf8encode("""Mein Behälter"" C:\Data\Mein Behälter\", 0, 1), 1);

[MBaas]

Looks good as werll:

Code: Select all

00000000: 22 4D 65 69 6E 20 42 65 68 C3 A4 6C 74 65 72 22   "Mein Behälter"
00000010: 20 43 3A 5C 44 61 74 61 5C 4D 65 69 6E 20 42 65    C:\Data\Mein Be
00000020: 68 C3 A4 6C 74 65 72 5C                           hälter\        

[admin]

Hmm. Are you sure that UltraEdit is showing the truth?

[MBaas]

It never lied to me (so far). i#ve attached the zipped file if you wanna check yourself.

[admin]

Yes, it's true. Can't explain it. Seems impossible.

What happens when you put more "ä" characters in the string? Any patterns?

[MBaas]

It's weird! I tried to write lots of "ä"s - they all were correct. Just my real use case (with the folder name "Gehälter") and the random repro seem to expose that behavior...

[admin]

It might be a different "ä" pasted from somewhere. Try to delete this character and type it again.

[MBaas]

Indeed - the folder name had "ä" in it (61 CC C9) which was faithfully reproduced when the file was created. It's many years old - I have no idea how it got in there. Possibly I used an inferior file mgr and for sure a different O/S. What a weird one - thanks for bearing with me and apologies for being so blind!