Shorteners self-test result about mcjson

[hydrobyte]

Hi,

Please, test the "main" branch for decimal separator awareness and other fixes. See InternalFloatToStr and InternalStrToFloat.

The "Shorteners" branch will be merged to "main" soon.

Regards,

[totyaxy]

I thought the shortener was a newer version. The trunk version with "," or "." only 1 test failed (save and load using files).

[hydrobyte]

Nice.
Could you debug it and report why this test failed?
Here I have two Windows PCs with Lazarus 2.2.0 and 2.3.0 and both have zero failures.

[totyaxy]

Hi, yes.

If any of them allowed then reasult test failed:

// load a Ansi file (especifying it)
M.LoadFromFile('test13-Ansi.json', false);
Result := Result and (M['ansi'].AsString = 'ăçüö');

// load a UTF-8 file
M.LoadFromFile('test13-UTF8.json');
Result := Result and (M['utf8'].AsString = 'ăçüö');

But not a surprise, as I see {$mode delphi} , and {$codepage cp1252} on the file header.

If I see these file with notepad++

1. test13-Ansi.json (force cp1252 codepage because I see garbage otherwise):
   {"ansi":"ãçüö"} (it's different as you see)

2. test13-UTF8.json (auto UTF8 detected correctly):
   {"utf8":"ãçüö"} (it's different too as you see)

Then I try to rewrite the "ă" to "ã" in the source, then I got message from the Lazarus: "save file cp1250 loose characters.."

So this is character encoding problem.

[hydrobyte]

Hi,

1. "test13-Ansi.json" is a ANSI file with CP-1252 characters. Here in my Notepad++ (via menu Encoding > Character sets > Western European > Windows-1252) I can see "{"ansi":"ãçüö"}" as expected.

2. I've set {$codepage cp1252} in PrjTestMcJSON.lpr because the .lpr file is not encoded in UTF-8. It opens correctly, it compiles and run OK.

3. Into your description I just see "ãçüö". There is no "ă" (items 1 and 2).

I think this error is related with your Windows (System) Codepage, that might be different from 1252. I just don't know why {$codepage cp1252} failed in order to prevent this CodePage misleading.

I'll explore changing my source code files to UTF-8, but I think this will broke Delphi (old versions) and Lazarus/FPC compatibility. I'll try changing just the .lpr file first to UTF-8 and improve Test13.

On the other hand, I think this error can't stop you from using McJSON if you work with .json files encoded in UTF-8.

Thanks for sharing. Stay tuned to next commits.

Regards,

[totyaxy]

Hi, I stayed, with the latest revision 94, the result of test 13 is still failed (Laz 2.2.4 and Laz 2.3.0 both same results)

[hydrobyte]

Hi,

I was revising Lazarus support to UTF-8, comparing it to Delphi pre-Unicode, and I think it is expected that Test13 fails with a Windows CodePage different from 1252.

I'll close this thread because it is in the wrong branch.

Please, create a new Issue within main branch and report your Windows CodePage.

Regards,