Browse Source

Detect icu4c error codes more precisely

Negative values are warnings, not errors; failing on them breals encoding when
it would otherwise work.

Unfortunately, these negative numbers are being converted into large positive
ones through the cgo bridge, so this becomes a slightly more complicated
check than necessary.
main
Nick Thomas 5 years ago
parent
commit
03c771153c
  1. 34
      c_bridge.c
  2. 8
      convert.go
  3. 30
      detect.go

34
c_bridge.c

@ -6,16 +6,16 @@
#include <unicode/ucnv.h>
// See description in c_bridge.h
const int detectCharset(void *detector,
void *input,
int input_len,
int *status,
MatchData *matchBuffer,
const int detectCharset(void *detector,
void *input,
int input_len,
int *status,
MatchData *matchBuffer,
int matchBufferSize) {
// Put input bytes in the detector.
ucsdet_setText((UCharsetDetector*)detector, (char*)input, input_len, status);
if (*status != U_ZERO_ERROR) {
if U_FAILURE(*status) {
return 0;
}
@ -25,7 +25,7 @@ const int detectCharset(void *detector,
// Perform analysis and return all guesses and their count.
bestGuesses = ucsdet_detectAll((UCharsetDetector*)detector, &matchCount, status);
if (*status != U_ZERO_ERROR) {
if U_FAILURE(*status) {
return 0;
}
@ -42,19 +42,19 @@ const int detectCharset(void *detector,
// Fill guessed encoding
bestGuessedCharset = ucsdet_getName(bestGuess, status);
if (*status != U_ZERO_ERROR) {
if U_FAILURE(*status) {
return 0;
}
// Fill guessed language
bestGuessedLanguage = ucsdet_getLanguage(bestGuess, status);
if (*status != U_ZERO_ERROR) {
if U_FAILURE(*status) {
return 0;
}
// Fill its confidence rating
int32_t conf = ucsdet_getConfidence(bestGuess, status);
if (*status != U_ZERO_ERROR) {
if U_FAILURE(*status) {
return 0;
}
@ -69,7 +69,7 @@ const int detectCharset(void *detector,
// See description in c_bridge.h
int convertToUtf16(const char *srcEncoding,
UChar *dest,
UChar *dest,
int32_t destCapacity,
const char *src,
int32_t srcLength,
@ -77,13 +77,13 @@ int convertToUtf16(const char *srcEncoding,
UConverter *conv;
conv = ucnv_open(srcEncoding, status);
if (*status != U_ZERO_ERROR) {
if U_FAILURE(*status) {
return 0;
}
/* Convert from original encoding to UTF-16 */
int len = ucnv_toUChars(conv, dest, destCapacity, src, srcLength, status);
if (*status != U_ZERO_ERROR) {
if U_FAILURE(*status) {
return 0;
}
@ -94,7 +94,7 @@ int convertToUtf16(const char *srcEncoding,
// See description in c_bridge.h
int convertFromUtf16(const char *destEncoding,
char *dest,
char *dest,
int32_t destCapacity,
const UChar *src,
int32_t srcLength,
@ -102,17 +102,17 @@ int convertFromUtf16(const char *destEncoding,
UConverter *conv;
conv = ucnv_open(destEncoding, status);
if (*status != U_ZERO_ERROR) {
if U_FAILURE(*status) {
return 0;
}
/* Convert from UTF-16 to destination encoding */
int len = ucnv_fromUChars(conv, dest, destCapacity, src, srcLength, status);
if (*status != U_ZERO_ERROR) {
if U_FAILURE(*status) {
return 0;
}
ucnv_close(conv);
return len;
}
}

8
convert.go

@ -14,7 +14,7 @@ const (
DefaultMaxTextSize = 1024 * 1024 // Default value for the max text length in conversion operations
utf8MaxCharSize = 4
utf16MaxCharSize = 4
)
)
var (
Utf8CString = C.CString("UTF-8")
@ -35,7 +35,7 @@ type CharsetConverter struct {
// are created in memory once and then used. 'maxTextSize' sets the size of these buffers.
// ICU library would return error if any processed text is longer than this parameter.
//
// NOTE:
// NOTE:
//
// UTF8 uses 1 to 4 bytes for each symbol.
// UTF16 uses 2 bytes to 4 bytes for each symbol.
@ -79,7 +79,7 @@ func (conv *CharsetConverter) ConvertToUtf8(input []byte, srcEncoding string) ([
C.int32_t(len(input)),
(*C.int)(unsafe.Pointer(&status)))
if status == U_ZERO_ERROR {
if isSuccess(status) {
nConvLen := C.convertFromUtf16(
Utf8CString,
(*C.char)(unsafe.Pointer(&conv.utf8Buffer[0])),
@ -88,7 +88,7 @@ func (conv *CharsetConverter) ConvertToUtf8(input []byte, srcEncoding string) ([
C.int32_t(convLen),
(*C.int)(unsafe.Pointer(&status)))
if status == U_ZERO_ERROR {
if isSuccess(status) {
resStr := conv.utf8Buffer[:nConvLen]
return ([]byte)(resStr), nil
}

30
detect.go

@ -11,9 +11,21 @@ import (
)
const (
U_ZERO_ERROR = 0 // ICU common constant error code which means that no error occured
MatchDataBufferSize = 25 // Size of the buffer for detection results (Max count of returned guesses per detect call)
)
U_ZERO_ERROR = 0 // ICU common constant error code which means that no error occured
U_ERROR_LIMIT = 0x7FFFFFFF // Dirty hack, negative error codes are are being turned into large positive ints
MatchDataBufferSize = 25 // Size of the buffer for detection results (Max count of returned guesses per detect call)
)
// Go implementation of the icu U_SUCCESS macro. Negative status codes are
// warnings, 0 is a success without warnings, > 0 is an error
func isSuccess(status int) bool {
return status <= U_ZERO_ERROR || status >= U_ERROR_LIMIT
}
// Go implementation of the icu U_FAILURE macro.
func isFailure(status int) bool {
return status > U_ZERO_ERROR && status < U_ERROR_LIMIT
}
// CharsetDetector provides ICU charset detection functionality.
type CharsetDetector struct {
@ -39,7 +51,7 @@ func NewCharsetDetector() (*CharsetDetector, error) {
det.ptr = C.ucsdet_open((*C.UErrorCode)(statusPtr))
if status != U_ZERO_ERROR {
if isFailure(status) {
return nil, fmt.Errorf("ICU Error code returned: %d", status)
}
@ -63,14 +75,14 @@ func (det *CharsetDetector) GuessCharset(input []byte) (matches []Match, err err
// Perform detection. Guess count is the number of matches returned.
// The matches themself are put in the result buffer
guessCount := C.detectCharset(
unsafe.Pointer(det.ptr),
unsafe.Pointer(&input[0]),
C.int(inputLen),
(*C.int)(unsafe.Pointer(&status)),
unsafe.Pointer(det.ptr),
unsafe.Pointer(&input[0]),
C.int(inputLen),
(*C.int)(unsafe.Pointer(&status)),
(*C.MatchData)(unsafe.Pointer(&det.resBuffer[0])),
C.int(MatchDataBufferSize))
if status == U_ZERO_ERROR {
if isSuccess(status) {
// Convert the returned number of entries from result buffer to a slice
// that will be returned
count := int(guessCount)

Loading…
Cancel
Save