GDAL failed to open a Chinese path FileGDB file

[liminlu0314]

GDAL failed to open a Chinese path FileGDB file. Test data are in the attachment, including three layers:

/测试
/poly
/test/polygons

I try to modify some of the source code, and now data with Chinese path can be opened, such as the attached"/测试". But“/poly”and“/test/polygons”still cannot be opened.

The following is my modified code in file FGdbUtils.cpp：

std::wstring StringToWString(const std::string& utf8string)
{
	char* pszUTF8String = NULL;
	if(CSLTestBoolean(CPLGetConfigOption("GDAL_FILENAME_IS_UTF8", "YES")))
		pszUTF8String = CPLStrdup(utf8string.c_str());
	else
		pszUTF8String = CPLRecode(utf8string.c_str(), CPL_ENC_LOCALE, CPL_ENC_UTF8);

    wchar_t* pszUTF16 = CPLRecodeToWChar( pszUTF8String, CPL_ENC_UTF8, CPL_ENC_UCS2);
    std::wstring utf16string = pszUTF16;
    CPLFree(pszUTF16);
	CPLFree(pszUTF8String);
    return utf16string;
}

std::string WStringToString(const std::wstring& utf16string)
{
    char* pszUTF8 = CPLRecodeFromWChar( utf16string.c_str(), CPL_ENC_UCS2, CPL_ENC_UTF8 );
    std::string utf8string = pszUTF8;
    CPLFree(pszUTF8);

	if(CSLTestBoolean(CPLGetConfigOption("GDAL_FILENAME_IS_UTF8", "YES")))
		return utf8string; 
	else
	{
		char* pszLocalString = CPLRecode(utf8string.c_str(), CPL_ENC_UTF8, CPL_ENC_LOCALE);
		utf8string = pszLocalString;
		CPLFree(pszLocalString);
		return utf8string; 
	}
}

[liminlu0314]

Test Data

[Even Rouault]

You are testing on Windows and with GDAL_FILENAME_IS_UTF8 = NO defined ? I'm a bit skeptical it helps since CPLRecode(xxxx, CPL_ENC_LOCALE, CPL_ENC_UTF8) is equivalent to CPLRecode(xxxx, CPL_ENC_ISO8859_1, CPL_ENC_UTF8) (see CPLRecodeStub() implementation). So that would mean that you pass a ISO8859-1 encoded filename ? I can't see how it can work with Chinese characters

On Linux, I don't need the patch, but indeed the FileGDB SDK seems to have issues with the poly and polygons layers :

$ ogrinfo 测试.gdb --debug on
FileGDB: Really opening 测试.gdb
FGDB: Error opening \poly. Skipping it (Invalid function arguments.)
FGDB: Error opening \test\polygons. Skipping it (Invalid function arguments.)
GDAL: GDALOpen(测试.gdb, this=0x10b9560) succeeds as FileGDB.
INFO: Open of `测试.gdb'
      using driver `FileGDB' successful.
OGR: GetLayerCount() = 1

1: 测试 (Point)
FileGDB: ref_count of 测试.gdb = 0 now
FileGDB: Really closing 测试.gdb now
GDAL: GDALClose(测试.gdb, this=0x10b9560)

I'm not sure something can be done regarding this, since the error comes from the SDK.

But for the record, on Linux, with the OpenFileGDB driver, I can list and read the 3 layers (note the -ro parameter that will trigger the OpenFileGDB driver):

$ ogrinfo 测试.gdb -ro
INFO: Open of `测试.gdb'
      using driver `OpenFileGDB' successful.
1: 测试 (Point)
2: poly (Multi Polygon)
3: polygons (Multi Polygon)

[liminlu0314]

Replying to rouault: Yes, I did the test on Windows and set the GDAL_FILENAME_IS_UTF8 = NO. After debugging I founded it worked. The filename is encoded according to CP936 instead of ISO8859-1. Thanks a lot. Best regards.

You are testing on Windows and with GDAL_FILENAME_IS_UTF8 = NO defined ? I'm a bit skeptical it helps since CPLRecode(xxxx, CPL_ENC_LOCALE, CPL_ENC_UTF8) is equivalent to CPLRecode(xxxx, CPL_ENC_ISO8859_1, CPL_ENC_UTF8) (see CPLRecodeStub() implementation). So that would mean that you pass a ISO8859-1 encoded filename ? I can't see how it can work with Chinese characters

On Linux, I don't need the patch, but indeed the FileGDB SDK seems to have issues with the poly and polygons layers :

$ ogrinfo 测试.gdb --debug on
FileGDB: Really opening 测试.gdb
FGDB: Error opening \poly. Skipping it (Invalid function arguments.)
FGDB: Error opening \test\polygons. Skipping it (Invalid function arguments.)
GDAL: GDALOpen(测试.gdb, this=0x10b9560) succeeds as FileGDB.
INFO: Open of `测试.gdb'
      using driver `FileGDB' successful.
OGR: GetLayerCount() = 1

1: 测试 (Point)
FileGDB: ref_count of 测试.gdb = 0 now
FileGDB: Really closing 测试.gdb now
GDAL: GDALClose(测试.gdb, this=0x10b9560)

I'm not sure something can be done regarding this, since the error comes from the SDK.

But for the record, on Linux, with the OpenFileGDB driver, I can list and read the 3 layers (note the -ro parameter that will trigger the OpenFileGDB driver):

$ ogrinfo 测试.gdb -ro
INFO: Open of `测试.gdb'
      using driver `OpenFileGDB' successful.
1: 测试 (Point)
2: poly (Multi Polygon)
3: polygons (Multi Polygon)