From 4001847a98f4b0323a85648dbe1723662a68c5d7 Mon Sep 17 00:00:00 2001 From: Alexander Belopolsky Date: Sat, 26 Feb 2011 01:02:56 +0000 Subject: [PATCH] PEP 7 conformance changes (whitespace only). --- Objects/unicodeobject.c | 634 ++++++++++++++++++++++------------------ 1 file changed, 350 insertions(+), 284 deletions(-) diff --git a/Objects/unicodeobject.c b/Objects/unicodeobject.c index 48ea0a2043e..19673859cf4 100644 --- a/Objects/unicodeobject.c +++ b/Objects/unicodeobject.c @@ -142,16 +142,18 @@ const unsigned char _Py_ascii_whitespace[] = { 0, 0, 0, 0, 0, 0, 0, 0 }; -static PyObject *unicode_encode_call_errorhandler(const char *errors, +static PyObject * +unicode_encode_call_errorhandler(const char *errors, PyObject **errorHandler,const char *encoding, const char *reason, const Py_UNICODE *unicode, Py_ssize_t size, PyObject **exceptionObject, Py_ssize_t startpos, Py_ssize_t endpos, Py_ssize_t *newpos); -static void raise_encode_exception(PyObject **exceptionObject, - const char *encoding, - const Py_UNICODE *unicode, Py_ssize_t size, - Py_ssize_t startpos, Py_ssize_t endpos, - const char *reason); +static void +raise_encode_exception(PyObject **exceptionObject, + const char *encoding, + const Py_UNICODE *unicode, Py_ssize_t size, + Py_ssize_t startpos, Py_ssize_t endpos, + const char *reason); /* Same for linebreaks */ static unsigned char ascii_linebreak[] = { @@ -223,7 +225,8 @@ static BLOOM_MASK bloom_linebreak; ((ch) < 128U ? ascii_linebreak[(ch)] : \ (BLOOM(bloom_linebreak, (ch)) && Py_UNICODE_ISLINEBREAK(ch))) -Py_LOCAL_INLINE(BLOOM_MASK) make_bloom_mask(Py_UNICODE* ptr, Py_ssize_t len) +Py_LOCAL_INLINE(BLOOM_MASK) +make_bloom_mask(Py_UNICODE* ptr, Py_ssize_t len) { /* calculate simple bloom-style bitmask for a given unicode string */ @@ -237,7 +240,8 @@ Py_LOCAL_INLINE(BLOOM_MASK) make_bloom_mask(Py_UNICODE* ptr, Py_ssize_t len) return mask; } -Py_LOCAL_INLINE(int) unicode_member(Py_UNICODE chr, Py_UNICODE* set, Py_ssize_t setlen) +Py_LOCAL_INLINE(int) +unicode_member(Py_UNICODE chr, Py_UNICODE* set, Py_ssize_t setlen) { Py_ssize_t i; @@ -253,9 +257,9 @@ Py_LOCAL_INLINE(int) unicode_member(Py_UNICODE chr, Py_UNICODE* set, Py_ssize_t /* --- Unicode Object ----------------------------------------------------- */ -static -int unicode_resize(register PyUnicodeObject *unicode, - Py_ssize_t length) +static int +unicode_resize(register PyUnicodeObject *unicode, + Py_ssize_t length) { void *oldstr; @@ -311,8 +315,8 @@ int unicode_resize(register PyUnicodeObject *unicode, */ -static -PyUnicodeObject *_PyUnicode_New(Py_ssize_t length) +static PyUnicodeObject * +_PyUnicode_New(Py_ssize_t length) { register PyUnicodeObject *unicode; @@ -383,8 +387,8 @@ PyUnicodeObject *_PyUnicode_New(Py_ssize_t length) return NULL; } -static -void unicode_dealloc(register PyUnicodeObject *unicode) +static void +unicode_dealloc(register PyUnicodeObject *unicode) { switch (PyUnicode_CHECK_INTERNED(unicode)) { case SSTATE_NOT_INTERNED: @@ -428,8 +432,8 @@ void unicode_dealloc(register PyUnicodeObject *unicode) } } -static -int _PyUnicode_Resize(PyUnicodeObject **unicode, Py_ssize_t length) +static int +_PyUnicode_Resize(PyUnicodeObject **unicode, Py_ssize_t length) { register PyUnicodeObject *v; @@ -464,13 +468,14 @@ int _PyUnicode_Resize(PyUnicodeObject **unicode, Py_ssize_t length) return unicode_resize(v, length); } -int PyUnicode_Resize(PyObject **unicode, Py_ssize_t length) +int +PyUnicode_Resize(PyObject **unicode, Py_ssize_t length) { return _PyUnicode_Resize((PyUnicodeObject **)unicode, length); } -PyObject *PyUnicode_FromUnicode(const Py_UNICODE *u, - Py_ssize_t size) +PyObject * +PyUnicode_FromUnicode(const Py_UNICODE *u, Py_ssize_t size) { PyUnicodeObject *unicode; @@ -511,7 +516,8 @@ PyObject *PyUnicode_FromUnicode(const Py_UNICODE *u, return (PyObject *)unicode; } -PyObject *PyUnicode_FromStringAndSize(const char *u, Py_ssize_t size) +PyObject * +PyUnicode_FromStringAndSize(const char *u, Py_ssize_t size) { PyUnicodeObject *unicode; @@ -558,7 +564,8 @@ PyObject *PyUnicode_FromStringAndSize(const char *u, Py_ssize_t size) return (PyObject *)unicode; } -PyObject *PyUnicode_FromString(const char *u) +PyObject * +PyUnicode_FromString(const char *u) { size_t size = strlen(u); if (size > PY_SSIZE_T_MAX) { @@ -580,8 +587,8 @@ PyObject *PyUnicode_FromString(const char *u) /* Here sizeof(wchar_t) is 4 but Py_UNICODE_SIZE == 2, so we need to convert from UTF32 to UTF16. */ -PyObject *PyUnicode_FromWideChar(register const wchar_t *w, - Py_ssize_t size) +PyObject * +PyUnicode_FromWideChar(register const wchar_t *w, Py_ssize_t size) { PyUnicodeObject *unicode; register Py_ssize_t i; @@ -631,8 +638,8 @@ PyObject *PyUnicode_FromWideChar(register const wchar_t *w, #else -PyObject *PyUnicode_FromWideChar(register const wchar_t *w, - Py_ssize_t size) +PyObject * +PyUnicode_FromWideChar(register const wchar_t *w, Py_ssize_t size) { PyUnicodeObject *unicode; @@ -1324,7 +1331,8 @@ PyUnicode_AsWideCharString(PyObject *unicode, #endif -PyObject *PyUnicode_FromOrdinal(int ordinal) +PyObject * +PyUnicode_FromOrdinal(int ordinal) { Py_UNICODE s[2]; @@ -1347,7 +1355,8 @@ PyObject *PyUnicode_FromOrdinal(int ordinal) return PyUnicode_FromUnicode(s, 1); } -PyObject *PyUnicode_FromObject(register PyObject *obj) +PyObject * +PyUnicode_FromObject(register PyObject *obj) { /* XXX Perhaps we should make this API an alias of PyObject_Str() instead ?! */ @@ -1367,9 +1376,10 @@ PyObject *PyUnicode_FromObject(register PyObject *obj) return NULL; } -PyObject *PyUnicode_FromEncodedObject(register PyObject *obj, - const char *encoding, - const char *errors) +PyObject * +PyUnicode_FromEncodedObject(register PyObject *obj, + const char *encoding, + const char *errors) { Py_buffer buffer; PyObject *v; @@ -1452,10 +1462,11 @@ normalize_encoding(const char *encoding, return 1; } -PyObject *PyUnicode_Decode(const char *s, - Py_ssize_t size, - const char *encoding, - const char *errors) +PyObject * +PyUnicode_Decode(const char *s, + Py_ssize_t size, + const char *encoding, + const char *errors) { PyObject *buffer = NULL, *unicode; Py_buffer info; @@ -1510,9 +1521,10 @@ PyObject *PyUnicode_Decode(const char *s, return NULL; } -PyObject *PyUnicode_AsDecodedObject(PyObject *unicode, - const char *encoding, - const char *errors) +PyObject * +PyUnicode_AsDecodedObject(PyObject *unicode, + const char *encoding, + const char *errors) { PyObject *v; @@ -1534,9 +1546,10 @@ PyObject *PyUnicode_AsDecodedObject(PyObject *unicode, return NULL; } -PyObject *PyUnicode_AsDecodedUnicode(PyObject *unicode, - const char *encoding, - const char *errors) +PyObject * +PyUnicode_AsDecodedUnicode(PyObject *unicode, + const char *encoding, + const char *errors) { PyObject *v; @@ -1565,10 +1578,11 @@ PyObject *PyUnicode_AsDecodedUnicode(PyObject *unicode, return NULL; } -PyObject *PyUnicode_Encode(const Py_UNICODE *s, - Py_ssize_t size, - const char *encoding, - const char *errors) +PyObject * +PyUnicode_Encode(const Py_UNICODE *s, + Py_ssize_t size, + const char *encoding, + const char *errors) { PyObject *v, *unicode; @@ -1580,9 +1594,10 @@ PyObject *PyUnicode_Encode(const Py_UNICODE *s, return v; } -PyObject *PyUnicode_AsEncodedObject(PyObject *unicode, - const char *encoding, - const char *errors) +PyObject * +PyUnicode_AsEncodedObject(PyObject *unicode, + const char *encoding, + const char *errors) { PyObject *v; @@ -1659,9 +1674,10 @@ PyUnicode_EncodeFSDefault(PyObject *unicode) #endif } -PyObject *PyUnicode_AsEncodedString(PyObject *unicode, - const char *encoding, - const char *errors) +PyObject * +PyUnicode_AsEncodedString(PyObject *unicode, + const char *encoding, + const char *errors) { PyObject *v; char lower[11]; /* Enough for any encoding shortcut */ @@ -1735,9 +1751,10 @@ PyObject *PyUnicode_AsEncodedString(PyObject *unicode, return NULL; } -PyObject *PyUnicode_AsEncodedUnicode(PyObject *unicode, - const char *encoding, - const char *errors) +PyObject * +PyUnicode_AsEncodedUnicode(PyObject *unicode, + const char *encoding, + const char *errors) { PyObject *v; @@ -1766,8 +1783,9 @@ PyObject *PyUnicode_AsEncodedUnicode(PyObject *unicode, return NULL; } -PyObject *_PyUnicode_AsDefaultEncodedString(PyObject *unicode, - const char *errors) +PyObject * +_PyUnicode_AsDefaultEncodedString(PyObject *unicode, + const char *errors) { PyObject *v = ((PyUnicodeObject *)unicode)->defenc; if (v) @@ -1933,7 +1951,8 @@ _PyUnicode_AsString(PyObject *unicode) return _PyUnicode_AsStringAndSize(unicode, NULL); } -Py_UNICODE *PyUnicode_AsUnicode(PyObject *unicode) +Py_UNICODE * +PyUnicode_AsUnicode(PyObject *unicode) { if (!PyUnicode_Check(unicode)) { PyErr_BadArgument(); @@ -1945,7 +1964,8 @@ Py_UNICODE *PyUnicode_AsUnicode(PyObject *unicode) return NULL; } -Py_ssize_t PyUnicode_GetSize(PyObject *unicode) +Py_ssize_t +PyUnicode_GetSize(PyObject *unicode) { if (!PyUnicode_Check(unicode)) { PyErr_BadArgument(); @@ -1957,7 +1977,8 @@ Py_ssize_t PyUnicode_GetSize(PyObject *unicode) return -1; } -const char *PyUnicode_GetDefaultEncoding(void) +const char * +PyUnicode_GetDefaultEncoding(void) { return "utf-8"; } @@ -1996,12 +2017,12 @@ onError: return 0 on success, -1 on error */ -static -int unicode_decode_call_errorhandler(const char *errors, PyObject **errorHandler, - const char *encoding, const char *reason, - const char **input, const char **inend, Py_ssize_t *startinpos, - Py_ssize_t *endinpos, PyObject **exceptionObject, const char **inptr, - PyUnicodeObject **output, Py_ssize_t *outpos, Py_UNICODE **outptr) +static int +unicode_decode_call_errorhandler(const char *errors, PyObject **errorHandler, + const char *encoding, const char *reason, + const char **input, const char **inend, Py_ssize_t *startinpos, + Py_ssize_t *endinpos, PyObject **exceptionObject, const char **inptr, + PyUnicodeObject **output, Py_ssize_t *outpos, Py_UNICODE **outptr) { static char *argparse = "O!n;decoding error handler must return (str, int) tuple"; @@ -2171,9 +2192,10 @@ char utf7_category[128] = { (directWS && (utf7_category[(c)] == 2)) || \ (directO && (utf7_category[(c)] == 1)))) -PyObject *PyUnicode_DecodeUTF7(const char *s, - Py_ssize_t size, - const char *errors) +PyObject * +PyUnicode_DecodeUTF7(const char *s, + Py_ssize_t size, + const char *errors) { return PyUnicode_DecodeUTF7Stateful(s, size, errors, NULL); } @@ -2185,10 +2207,11 @@ PyObject *PyUnicode_DecodeUTF7(const char *s, * all the shift state (seen bits, number of bits seen, high * surrogate). */ -PyObject *PyUnicode_DecodeUTF7Stateful(const char *s, - Py_ssize_t size, - const char *errors, - Py_ssize_t *consumed) +PyObject * +PyUnicode_DecodeUTF7Stateful(const char *s, + Py_ssize_t size, + const char *errors, + Py_ssize_t *consumed) { const char *starts = s; Py_ssize_t startinpos; @@ -2375,11 +2398,12 @@ utf7Error: } -PyObject *PyUnicode_EncodeUTF7(const Py_UNICODE *s, - Py_ssize_t size, - int base64SetO, - int base64WhiteSpace, - const char *errors) +PyObject * +PyUnicode_EncodeUTF7(const Py_UNICODE *s, + Py_ssize_t size, + int base64SetO, + int base64WhiteSpace, + const char *errors) { PyObject *v; /* It might be possible to tighten this worst case */ @@ -2500,9 +2524,10 @@ char utf8_code_length[256] = { 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0-F4 + F5-FF */ }; -PyObject *PyUnicode_DecodeUTF8(const char *s, - Py_ssize_t size, - const char *errors) +PyObject * +PyUnicode_DecodeUTF8(const char *s, + Py_ssize_t size, + const char *errors) { return PyUnicode_DecodeUTF8Stateful(s, size, errors, NULL); } @@ -2520,10 +2545,11 @@ PyObject *PyUnicode_DecodeUTF8(const char *s, # error C 'long' size should be either 4 or 8! #endif -PyObject *PyUnicode_DecodeUTF8Stateful(const char *s, - Py_ssize_t size, - const char *errors, - Py_ssize_t *consumed) +PyObject * +PyUnicode_DecodeUTF8Stateful(const char *s, + Py_ssize_t size, + const char *errors, + Py_ssize_t *consumed) { const char *starts = s; int n; @@ -3029,7 +3055,8 @@ PyUnicode_EncodeUTF8(const Py_UNICODE *s, #undef MAX_SHORT_UNICHARS } -PyObject *PyUnicode_AsUTF8String(PyObject *unicode) +PyObject * +PyUnicode_AsUTF8String(PyObject *unicode) { if (!PyUnicode_Check(unicode)) { PyErr_BadArgument(); @@ -3303,7 +3330,8 @@ PyUnicode_EncodeUTF32(const Py_UNICODE *s, #undef STORECHAR } -PyObject *PyUnicode_AsUTF32String(PyObject *unicode) +PyObject * +PyUnicode_AsUTF32String(PyObject *unicode) { if (!PyUnicode_Check(unicode)) { PyErr_BadArgument(); @@ -3693,7 +3721,8 @@ PyUnicode_EncodeUTF16(const Py_UNICODE *s, #undef STORECHAR } -PyObject *PyUnicode_AsUTF16String(PyObject *unicode) +PyObject * +PyUnicode_AsUTF16String(PyObject *unicode) { if (!PyUnicode_Check(unicode)) { PyErr_BadArgument(); @@ -3709,9 +3738,10 @@ PyObject *PyUnicode_AsUTF16String(PyObject *unicode) static _PyUnicode_Name_CAPI *ucnhash_CAPI = NULL; -PyObject *PyUnicode_DecodeUnicodeEscape(const char *s, - Py_ssize_t size, - const char *errors) +PyObject * +PyUnicode_DecodeUnicodeEscape(const char *s, + Py_ssize_t size, + const char *errors) { const char *starts = s; Py_ssize_t startinpos; @@ -3967,8 +3997,9 @@ Py_LOCAL_INLINE(const Py_UNICODE *) findchar(const Py_UNICODE *s, static const char *hexdigits = "0123456789abcdef"; -PyObject *PyUnicode_EncodeUnicodeEscape(const Py_UNICODE *s, - Py_ssize_t size) +PyObject * +PyUnicode_EncodeUnicodeEscape(const Py_UNICODE *s, + Py_ssize_t size) { PyObject *repr; char *p; @@ -4108,7 +4139,8 @@ PyObject *PyUnicode_EncodeUnicodeEscape(const Py_UNICODE *s, return repr; } -PyObject *PyUnicode_AsUnicodeEscapeString(PyObject *unicode) +PyObject * +PyUnicode_AsUnicodeEscapeString(PyObject *unicode) { PyObject *s; if (!PyUnicode_Check(unicode)) { @@ -4122,9 +4154,10 @@ PyObject *PyUnicode_AsUnicodeEscapeString(PyObject *unicode) /* --- Raw Unicode Escape Codec ------------------------------------------- */ -PyObject *PyUnicode_DecodeRawUnicodeEscape(const char *s, - Py_ssize_t size, - const char *errors) +PyObject * +PyUnicode_DecodeRawUnicodeEscape(const char *s, + Py_ssize_t size, + const char *errors) { const char *starts = s; Py_ssize_t startinpos; @@ -4239,8 +4272,9 @@ PyObject *PyUnicode_DecodeRawUnicodeEscape(const char *s, return NULL; } -PyObject *PyUnicode_EncodeRawUnicodeEscape(const Py_UNICODE *s, - Py_ssize_t size) +PyObject * +PyUnicode_EncodeRawUnicodeEscape(const Py_UNICODE *s, + Py_ssize_t size) { PyObject *repr; char *p; @@ -4327,7 +4361,8 @@ PyObject *PyUnicode_EncodeRawUnicodeEscape(const Py_UNICODE *s, return repr; } -PyObject *PyUnicode_AsRawUnicodeEscapeString(PyObject *unicode) +PyObject * +PyUnicode_AsRawUnicodeEscapeString(PyObject *unicode) { PyObject *s; if (!PyUnicode_Check(unicode)) { @@ -4342,9 +4377,10 @@ PyObject *PyUnicode_AsRawUnicodeEscapeString(PyObject *unicode) /* --- Unicode Internal Codec ------------------------------------------- */ -PyObject *_PyUnicode_DecodeUnicodeInternal(const char *s, - Py_ssize_t size, - const char *errors) +PyObject * +_PyUnicode_DecodeUnicodeInternal(const char *s, + Py_ssize_t size, + const char *errors) { const char *starts = s; Py_ssize_t startinpos; @@ -4420,9 +4456,10 @@ PyObject *_PyUnicode_DecodeUnicodeInternal(const char *s, /* --- Latin-1 Codec ------------------------------------------------------ */ -PyObject *PyUnicode_DecodeLatin1(const char *s, - Py_ssize_t size, - const char *errors) +PyObject * +PyUnicode_DecodeLatin1(const char *s, + Py_ssize_t size, + const char *errors) { PyUnicodeObject *v; Py_UNICODE *p; @@ -4462,11 +4499,12 @@ PyObject *PyUnicode_DecodeLatin1(const char *s, } /* create or adjust a UnicodeEncodeError */ -static void make_encode_exception(PyObject **exceptionObject, - const char *encoding, - const Py_UNICODE *unicode, Py_ssize_t size, - Py_ssize_t startpos, Py_ssize_t endpos, - const char *reason) +static void +make_encode_exception(PyObject **exceptionObject, + const char *encoding, + const Py_UNICODE *unicode, Py_ssize_t size, + Py_ssize_t startpos, Py_ssize_t endpos, + const char *reason) { if (*exceptionObject == NULL) { *exceptionObject = PyUnicodeEncodeError_Create( @@ -4487,11 +4525,12 @@ static void make_encode_exception(PyObject **exceptionObject, } /* raises a UnicodeEncodeError */ -static void raise_encode_exception(PyObject **exceptionObject, - const char *encoding, - const Py_UNICODE *unicode, Py_ssize_t size, - Py_ssize_t startpos, Py_ssize_t endpos, - const char *reason) +static void +raise_encode_exception(PyObject **exceptionObject, + const char *encoding, + const Py_UNICODE *unicode, Py_ssize_t size, + Py_ssize_t startpos, Py_ssize_t endpos, + const char *reason) { make_encode_exception(exceptionObject, encoding, unicode, size, startpos, endpos, reason); @@ -4503,12 +4542,13 @@ static void raise_encode_exception(PyObject **exceptionObject, build arguments, call the callback and check the arguments, put the result into newpos and return the replacement string, which has to be freed by the caller */ -static PyObject *unicode_encode_call_errorhandler(const char *errors, - PyObject **errorHandler, - const char *encoding, const char *reason, - const Py_UNICODE *unicode, Py_ssize_t size, PyObject **exceptionObject, - Py_ssize_t startpos, Py_ssize_t endpos, - Py_ssize_t *newpos) +static PyObject * +unicode_encode_call_errorhandler(const char *errors, + PyObject **errorHandler, + const char *encoding, const char *reason, + const Py_UNICODE *unicode, Py_ssize_t size, PyObject **exceptionObject, + Py_ssize_t startpos, Py_ssize_t endpos, + Py_ssize_t *newpos) { static char *argparse = "On;encoding error handler must return (str/bytes, int) tuple"; @@ -4557,10 +4597,11 @@ static PyObject *unicode_encode_call_errorhandler(const char *errors, return resunicode; } -static PyObject *unicode_encode_ucs1(const Py_UNICODE *p, - Py_ssize_t size, - const char *errors, - int limit) +static PyObject * +unicode_encode_ucs1(const Py_UNICODE *p, + Py_ssize_t size, + const char *errors, + int limit) { /* output object */ PyObject *res; @@ -4753,14 +4794,16 @@ static PyObject *unicode_encode_ucs1(const Py_UNICODE *p, return NULL; } -PyObject *PyUnicode_EncodeLatin1(const Py_UNICODE *p, - Py_ssize_t size, - const char *errors) +PyObject * +PyUnicode_EncodeLatin1(const Py_UNICODE *p, + Py_ssize_t size, + const char *errors) { return unicode_encode_ucs1(p, size, errors, 256); } -PyObject *PyUnicode_AsLatin1String(PyObject *unicode) +PyObject * +PyUnicode_AsLatin1String(PyObject *unicode) { if (!PyUnicode_Check(unicode)) { PyErr_BadArgument(); @@ -4773,9 +4816,10 @@ PyObject *PyUnicode_AsLatin1String(PyObject *unicode) /* --- 7-bit ASCII Codec -------------------------------------------------- */ -PyObject *PyUnicode_DecodeASCII(const char *s, - Py_ssize_t size, - const char *errors) +PyObject * +PyUnicode_DecodeASCII(const char *s, + Py_ssize_t size, + const char *errors) { const char *starts = s; PyUnicodeObject *v; @@ -4832,14 +4876,16 @@ PyObject *PyUnicode_DecodeASCII(const char *s, return NULL; } -PyObject *PyUnicode_EncodeASCII(const Py_UNICODE *p, - Py_ssize_t size, - const char *errors) +PyObject * +PyUnicode_EncodeASCII(const Py_UNICODE *p, + Py_ssize_t size, + const char *errors) { return unicode_encode_ucs1(p, size, errors, 128); } -PyObject *PyUnicode_AsASCIIString(PyObject *unicode) +PyObject * +PyUnicode_AsASCIIString(PyObject *unicode) { if (!PyUnicode_Check(unicode)) { PyErr_BadArgument(); @@ -4863,7 +4909,8 @@ PyObject *PyUnicode_AsASCIIString(PyObject *unicode) b) IsDBCSLeadByte (probably) does not work for non-DBCS multi-byte encodings, see IsDBCSLeadByteEx documentation. */ -static int is_dbcs_lead_byte(const char *s, int offset) +static int +is_dbcs_lead_byte(const char *s, int offset) { const char *curr = s + offset; @@ -4878,11 +4925,12 @@ static int is_dbcs_lead_byte(const char *s, int offset) * Decode MBCS string into unicode object. If 'final' is set, converts * trailing lead-byte too. Returns consumed size if succeed, -1 otherwise. */ -static int decode_mbcs(PyUnicodeObject **v, - const char *s, /* MBCS string */ - int size, /* sizeof MBCS string */ - int final, - const char *errors) +static int +decode_mbcs(PyUnicodeObject **v, + const char *s, /* MBCS string */ + int size, /* sizeof MBCS string */ + int final, + const char *errors) { Py_UNICODE *p; Py_ssize_t n; @@ -4961,10 +5009,11 @@ mbcs_decode_error: return -1; } -PyObject *PyUnicode_DecodeMBCSStateful(const char *s, - Py_ssize_t size, - const char *errors, - Py_ssize_t *consumed) +PyObject * +PyUnicode_DecodeMBCSStateful(const char *s, + Py_ssize_t size, + const char *errors, + Py_ssize_t *consumed) { PyUnicodeObject *v = NULL; int done; @@ -4999,9 +5048,10 @@ PyObject *PyUnicode_DecodeMBCSStateful(const char *s, return (PyObject *)v; } -PyObject *PyUnicode_DecodeMBCS(const char *s, - Py_ssize_t size, - const char *errors) +PyObject * +PyUnicode_DecodeMBCS(const char *s, + Py_ssize_t size, + const char *errors) { return PyUnicode_DecodeMBCSStateful(s, size, errors, NULL); } @@ -5010,10 +5060,11 @@ PyObject *PyUnicode_DecodeMBCS(const char *s, * Convert unicode into string object (MBCS). * Returns 0 if succeed, -1 otherwise. */ -static int encode_mbcs(PyObject **repr, - const Py_UNICODE *p, /* unicode */ - int size, /* size of unicode */ - const char* errors) +static int +encode_mbcs(PyObject **repr, + const Py_UNICODE *p, /* unicode */ + int size, /* size of unicode */ + const char* errors) { BOOL usedDefaultChar = FALSE; BOOL *pusedDefaultChar; @@ -5086,9 +5137,10 @@ mbcs_encode_error: return -1; } -PyObject *PyUnicode_EncodeMBCS(const Py_UNICODE *p, - Py_ssize_t size, - const char *errors) +PyObject * +PyUnicode_EncodeMBCS(const Py_UNICODE *p, + Py_ssize_t size, + const char *errors) { PyObject *repr = NULL; int ret; @@ -5117,7 +5169,8 @@ PyObject *PyUnicode_EncodeMBCS(const Py_UNICODE *p, return repr; } -PyObject *PyUnicode_AsMBCSString(PyObject *unicode) +PyObject * +PyUnicode_AsMBCSString(PyObject *unicode) { if (!PyUnicode_Check(unicode)) { PyErr_BadArgument(); @@ -5134,10 +5187,11 @@ PyObject *PyUnicode_AsMBCSString(PyObject *unicode) /* --- Character Mapping Codec -------------------------------------------- */ -PyObject *PyUnicode_DecodeCharmap(const char *s, - Py_ssize_t size, - PyObject *mapping, - const char *errors) +PyObject * +PyUnicode_DecodeCharmap(const char *s, + Py_ssize_t size, + PyObject *mapping, + const char *errors) { const char *starts = s; Py_ssize_t startinpos; @@ -5297,7 +5351,7 @@ PyObject *PyUnicode_DecodeCharmap(const char *s, /* Charmap encoding: the lookup table */ -struct encoding_map{ +struct encoding_map { PyObject_HEAD unsigned char level1[32]; int count2, count3; @@ -5511,7 +5565,8 @@ encoding_map_lookup(Py_UNICODE c, PyObject *mapping) /* Lookup the character ch in the mapping. If the character can't be found, Py_None is returned (or NULL, if another error occurred). */ -static PyObject *charmapencode_lookup(Py_UNICODE c, PyObject *mapping) +static PyObject * +charmapencode_lookup(Py_UNICODE c, PyObject *mapping) { PyObject *w = PyLong_FromLong((long)c); PyObject *x; @@ -5568,16 +5623,16 @@ charmapencode_resize(PyObject **outobj, Py_ssize_t *outpos, Py_ssize_t requireds typedef enum charmapencode_result { enc_SUCCESS, enc_FAILED, enc_EXCEPTION -}charmapencode_result; +} charmapencode_result; /* lookup the character, put the result in the output string and adjust various state variables. Resize the output bytes object if not enough space is available. Return a new reference to the object that was put in the output buffer, or Py_None, if the mapping was undefined (in which case no character was written) or NULL, if a reallocation error occurred. The caller must decref the result */ -static -charmapencode_result charmapencode_output(Py_UNICODE c, PyObject *mapping, - PyObject **outobj, Py_ssize_t *outpos) +static charmapencode_result +charmapencode_output(Py_UNICODE c, PyObject *mapping, + PyObject **outobj, Py_ssize_t *outpos) { PyObject *rep; char *outstart; @@ -5633,8 +5688,8 @@ charmapencode_result charmapencode_output(Py_UNICODE c, PyObject *mapping, /* handle an error in PyUnicode_EncodeCharmap Return 0 on success, -1 on error */ -static -int charmap_encoding_error( +static int +charmap_encoding_error( const Py_UNICODE *p, Py_ssize_t size, Py_ssize_t *inpos, PyObject *mapping, PyObject **exceptionObject, int *known_errorHandler, PyObject **errorHandler, const char *errors, @@ -5768,10 +5823,11 @@ int charmap_encoding_error( return 0; } -PyObject *PyUnicode_EncodeCharmap(const Py_UNICODE *p, - Py_ssize_t size, - PyObject *mapping, - const char *errors) +PyObject * +PyUnicode_EncodeCharmap(const Py_UNICODE *p, + Py_ssize_t size, + PyObject *mapping, + const char *errors) { /* output object */ PyObject *res = NULL; @@ -5832,8 +5888,9 @@ PyObject *PyUnicode_EncodeCharmap(const Py_UNICODE *p, return NULL; } -PyObject *PyUnicode_AsCharmapString(PyObject *unicode, - PyObject *mapping) +PyObject * +PyUnicode_AsCharmapString(PyObject *unicode, + PyObject *mapping) { if (!PyUnicode_Check(unicode) || mapping == NULL) { PyErr_BadArgument(); @@ -5846,10 +5903,11 @@ PyObject *PyUnicode_AsCharmapString(PyObject *unicode, } /* create or adjust a UnicodeTranslateError */ -static void make_translate_exception(PyObject **exceptionObject, - const Py_UNICODE *unicode, Py_ssize_t size, - Py_ssize_t startpos, Py_ssize_t endpos, - const char *reason) +static void +make_translate_exception(PyObject **exceptionObject, + const Py_UNICODE *unicode, Py_ssize_t size, + Py_ssize_t startpos, Py_ssize_t endpos, + const char *reason) { if (*exceptionObject == NULL) { *exceptionObject = PyUnicodeTranslateError_Create( @@ -5870,10 +5928,11 @@ static void make_translate_exception(PyObject **exceptionObject, } /* raises a UnicodeTranslateError */ -static void raise_translate_exception(PyObject **exceptionObject, - const Py_UNICODE *unicode, Py_ssize_t size, - Py_ssize_t startpos, Py_ssize_t endpos, - const char *reason) +static void +raise_translate_exception(PyObject **exceptionObject, + const Py_UNICODE *unicode, Py_ssize_t size, + Py_ssize_t startpos, Py_ssize_t endpos, + const char *reason) { make_translate_exception(exceptionObject, unicode, size, startpos, endpos, reason); @@ -5885,12 +5944,13 @@ static void raise_translate_exception(PyObject **exceptionObject, build arguments, call the callback and check the arguments, put the result into newpos and return the replacement string, which has to be freed by the caller */ -static PyObject *unicode_translate_call_errorhandler(const char *errors, - PyObject **errorHandler, - const char *reason, - const Py_UNICODE *unicode, Py_ssize_t size, PyObject **exceptionObject, - Py_ssize_t startpos, Py_ssize_t endpos, - Py_ssize_t *newpos) +static PyObject * +unicode_translate_call_errorhandler(const char *errors, + PyObject **errorHandler, + const char *reason, + const Py_UNICODE *unicode, Py_ssize_t size, PyObject **exceptionObject, + Py_ssize_t startpos, Py_ssize_t endpos, + Py_ssize_t *newpos) { static char *argparse = "O!n;translating error handler must return (str, int) tuple"; @@ -5940,8 +6000,8 @@ static PyObject *unicode_translate_call_errorhandler(const char *errors, /* Lookup the character ch in the mapping and put the result in result, which must be decrefed by the caller. Return 0 on success, -1 on error */ -static -int charmaptranslate_lookup(Py_UNICODE c, PyObject *mapping, PyObject **result) +static int +charmaptranslate_lookup(Py_UNICODE c, PyObject *mapping, PyObject **result) { PyObject *w = PyLong_FromLong((long)c); PyObject *x; @@ -5990,8 +6050,8 @@ int charmaptranslate_lookup(Py_UNICODE c, PyObject *mapping, PyObject **result) /* ensure that *outobj is at least requiredsize characters long, if not reallocate and adjust various state variables. Return 0 on success, -1 on error */ -static -int charmaptranslate_makespace(PyObject **outobj, Py_UNICODE **outp, +static int +charmaptranslate_makespace(PyObject **outobj, Py_UNICODE **outp, Py_ssize_t requiredsize) { Py_ssize_t oldsize = PyUnicode_GET_SIZE(*outobj); @@ -6013,10 +6073,10 @@ int charmaptranslate_makespace(PyObject **outobj, Py_UNICODE **outp, undefined (in which case no character was written). The called must decref result. Return 0 on success, -1 on error. */ -static -int charmaptranslate_output(const Py_UNICODE *startinp, const Py_UNICODE *curinp, - Py_ssize_t insize, PyObject *mapping, PyObject **outobj, Py_UNICODE **outp, - PyObject **res) +static int +charmaptranslate_output(const Py_UNICODE *startinp, const Py_UNICODE *curinp, + Py_ssize_t insize, PyObject *mapping, PyObject **outobj, Py_UNICODE **outp, + PyObject **res) { if (charmaptranslate_lookup(*curinp, mapping, res)) return -1; @@ -6052,10 +6112,11 @@ int charmaptranslate_output(const Py_UNICODE *startinp, const Py_UNICODE *curinp return 0; } -PyObject *PyUnicode_TranslateCharmap(const Py_UNICODE *p, - Py_ssize_t size, - PyObject *mapping, - const char *errors) +PyObject * +PyUnicode_TranslateCharmap(const Py_UNICODE *p, + Py_ssize_t size, + PyObject *mapping, + const char *errors) { /* output object */ PyObject *res = NULL; @@ -6194,9 +6255,10 @@ PyObject *PyUnicode_TranslateCharmap(const Py_UNICODE *p, return NULL; } -PyObject *PyUnicode_Translate(PyObject *str, - PyObject *mapping, - const char *errors) +PyObject * +PyUnicode_Translate(PyObject *str, + PyObject *mapping, + const char *errors) { PyObject *result; @@ -6241,10 +6303,11 @@ PyUnicode_TransformDecimalToASCII(Py_UNICODE *s, } /* --- Decimal Encoder ---------------------------------------------------- */ -int PyUnicode_EncodeDecimal(Py_UNICODE *s, - Py_ssize_t length, - char *output, - const char *errors) +int +PyUnicode_EncodeDecimal(Py_UNICODE *s, + Py_ssize_t length, + char *output, + const char *errors) { Py_UNICODE *p, *end; PyObject *errorHandler = NULL; @@ -6405,10 +6468,11 @@ int PyUnicode_EncodeDecimal(Py_UNICODE *s, start = 0; \ } -Py_ssize_t PyUnicode_Count(PyObject *str, - PyObject *substr, - Py_ssize_t start, - Py_ssize_t end) +Py_ssize_t +PyUnicode_Count(PyObject *str, + PyObject *substr, + Py_ssize_t start, + Py_ssize_t end) { Py_ssize_t result; PyUnicodeObject* str_obj; @@ -6435,11 +6499,12 @@ Py_ssize_t PyUnicode_Count(PyObject *str, return result; } -Py_ssize_t PyUnicode_Find(PyObject *str, - PyObject *sub, - Py_ssize_t start, - Py_ssize_t end, - int direction) +Py_ssize_t +PyUnicode_Find(PyObject *str, + PyObject *sub, + Py_ssize_t start, + Py_ssize_t end, + int direction) { Py_ssize_t result; @@ -6471,12 +6536,12 @@ Py_ssize_t PyUnicode_Find(PyObject *str, return result; } -static -int tailmatch(PyUnicodeObject *self, - PyUnicodeObject *substring, - Py_ssize_t start, - Py_ssize_t end, - int direction) +static int +tailmatch(PyUnicodeObject *self, + PyUnicodeObject *substring, + Py_ssize_t start, + Py_ssize_t end, + int direction) { if (substring->length == 0) return 1; @@ -6497,11 +6562,12 @@ int tailmatch(PyUnicodeObject *self, return 0; } -Py_ssize_t PyUnicode_Tailmatch(PyObject *str, - PyObject *substr, - Py_ssize_t start, - Py_ssize_t end, - int direction) +Py_ssize_t +PyUnicode_Tailmatch(PyObject *str, + PyObject *substr, + Py_ssize_t start, + Py_ssize_t end, + int direction) { Py_ssize_t result; @@ -6525,9 +6591,9 @@ Py_ssize_t PyUnicode_Tailmatch(PyObject *str, /* Apply fixfct filter to the Unicode object self and return a reference to the modified object */ -static -PyObject *fixup(PyUnicodeObject *self, - int (*fixfct)(PyUnicodeObject *s)) +static PyObject * +fixup(PyUnicodeObject *self, + int (*fixfct)(PyUnicodeObject *s)) { PyUnicodeObject *u; @@ -6549,8 +6615,8 @@ PyObject *fixup(PyUnicodeObject *self, return (PyObject*) u; } -static -int fixupper(PyUnicodeObject *self) +static int +fixupper(PyUnicodeObject *self) { Py_ssize_t len = self->length; Py_UNICODE *s = self->str; @@ -6570,8 +6636,8 @@ int fixupper(PyUnicodeObject *self) return status; } -static -int fixlower(PyUnicodeObject *self) +static int +fixlower(PyUnicodeObject *self) { Py_ssize_t len = self->length; Py_UNICODE *s = self->str; @@ -6591,8 +6657,8 @@ int fixlower(PyUnicodeObject *self) return status; } -static -int fixswapcase(PyUnicodeObject *self) +static int +fixswapcase(PyUnicodeObject *self) { Py_ssize_t len = self->length; Py_UNICODE *s = self->str; @@ -6612,8 +6678,8 @@ int fixswapcase(PyUnicodeObject *self) return status; } -static -int fixcapitalize(PyUnicodeObject *self) +static int +fixcapitalize(PyUnicodeObject *self) { Py_ssize_t len = self->length; Py_UNICODE *s = self->str; @@ -6636,8 +6702,8 @@ int fixcapitalize(PyUnicodeObject *self) return status; } -static -int fixtitle(PyUnicodeObject *self) +static int +fixtitle(PyUnicodeObject *self) { register Py_UNICODE *p = PyUnicode_AS_UNICODE(self); register Py_UNICODE *e; @@ -6787,11 +6853,11 @@ PyUnicode_Join(PyObject *separator, PyObject *seq) return NULL; } -static -PyUnicodeObject *pad(PyUnicodeObject *self, - Py_ssize_t left, - Py_ssize_t right, - Py_UNICODE fill) +static PyUnicodeObject * +pad(PyUnicodeObject *self, + Py_ssize_t left, + Py_ssize_t right, + Py_UNICODE fill) { PyUnicodeObject *u; @@ -6822,7 +6888,8 @@ PyUnicodeObject *pad(PyUnicodeObject *self, return u; } -PyObject *PyUnicode_Splitlines(PyObject *string, int keepends) +PyObject * +PyUnicode_Splitlines(PyObject *string, int keepends) { PyObject *list; @@ -6838,10 +6905,10 @@ PyObject *PyUnicode_Splitlines(PyObject *string, int keepends) return list; } -static -PyObject *split(PyUnicodeObject *self, - PyUnicodeObject *substring, - Py_ssize_t maxcount) +static PyObject * +split(PyUnicodeObject *self, + PyUnicodeObject *substring, + Py_ssize_t maxcount) { if (maxcount < 0) maxcount = PY_SSIZE_T_MAX; @@ -6858,10 +6925,10 @@ PyObject *split(PyUnicodeObject *self, ); } -static -PyObject *rsplit(PyUnicodeObject *self, - PyUnicodeObject *substring, - Py_ssize_t maxcount) +static PyObject * +rsplit(PyUnicodeObject *self, + PyUnicodeObject *substring, + Py_ssize_t maxcount) { if (maxcount < 0) maxcount = PY_SSIZE_T_MAX; @@ -6878,11 +6945,11 @@ PyObject *rsplit(PyUnicodeObject *self, ); } -static -PyObject *replace(PyUnicodeObject *self, - PyUnicodeObject *str1, - PyUnicodeObject *str2, - Py_ssize_t maxcount) +static PyObject * +replace(PyUnicodeObject *self, + PyUnicodeObject *str1, + PyUnicodeObject *str2, + Py_ssize_t maxcount) { PyUnicodeObject *u; @@ -7216,8 +7283,8 @@ unicode_compare(PyUnicodeObject *str1, PyUnicodeObject *str2) #endif -int PyUnicode_Compare(PyObject *left, - PyObject *right) +int +PyUnicode_Compare(PyObject *left, PyObject *right) { if (PyUnicode_Check(left) && PyUnicode_Check(right)) return unicode_compare((PyUnicodeObject *)left, @@ -7253,9 +7320,8 @@ PyUnicode_CompareWithASCIIString(PyObject* uni, const char* str) #define TEST_COND(cond) \ ((cond) ? Py_True : Py_False) -PyObject *PyUnicode_RichCompare(PyObject *left, - PyObject *right, - int op) +PyObject * +PyUnicode_RichCompare(PyObject *left, PyObject *right, int op) { int result; @@ -7310,8 +7376,8 @@ PyObject *PyUnicode_RichCompare(PyObject *left, return Py_NotImplemented; } -int PyUnicode_Contains(PyObject *container, - PyObject *element) +int +PyUnicode_Contains(PyObject *container, PyObject *element) { PyObject *str, *sub; int result; @@ -7341,8 +7407,8 @@ int PyUnicode_Contains(PyObject *container, /* Concat to string or Unicode object giving a new Unicode object. */ -PyObject *PyUnicode_Concat(PyObject *left, - PyObject *right) +PyObject * +PyUnicode_Concat(PyObject *left, PyObject *right) { PyUnicodeObject *u = NULL, *v = NULL, *w; @@ -8258,10 +8324,11 @@ unicode_repeat(PyUnicodeObject *str, Py_ssize_t len) return (PyObject*) u; } -PyObject *PyUnicode_Replace(PyObject *obj, - PyObject *subobj, - PyObject *replobj, - Py_ssize_t maxcount) +PyObject * +PyUnicode_Replace(PyObject *obj, + PyObject *subobj, + PyObject *replobj, + Py_ssize_t maxcount) { PyObject *self; PyObject *str1; @@ -8325,8 +8392,8 @@ unicode_replace(PyUnicodeObject *self, PyObject *args) return result; } -static -PyObject *unicode_repr(PyObject *unicode) +static PyObject * +unicode_repr(PyObject *unicode) { PyObject *repr; Py_UNICODE *p; @@ -8559,9 +8626,8 @@ unicode_rjust(PyUnicodeObject *self, PyObject *args) return (PyObject*) pad(self, width - self->length, 0, fillchar); } -PyObject *PyUnicode_Split(PyObject *s, - PyObject *sep, - Py_ssize_t maxsplit) +PyObject * +PyUnicode_Split(PyObject *s, PyObject *sep, Py_ssize_t maxsplit) { PyObject *result; @@ -8690,9 +8756,8 @@ unicode_rpartition(PyUnicodeObject *self, PyObject *separator) return PyUnicode_RPartition((PyObject *)self, separator); } -PyObject *PyUnicode_RSplit(PyObject *s, - PyObject *sep, - Py_ssize_t maxsplit) +PyObject * +PyUnicode_RSplit(PyObject *s, PyObject *sep, Py_ssize_t maxsplit) { PyObject *result; @@ -9384,8 +9449,8 @@ formatchar(Py_UNICODE *buf, */ #define FORMATBUFLEN (size_t)10 -PyObject *PyUnicode_Format(PyObject *format, - PyObject *args) +PyObject * +PyUnicode_Format(PyObject *format, PyObject *args) { Py_UNICODE *fmt, *res; Py_ssize_t fmtcnt, rescnt, reslen, arglen, argidx; @@ -10076,7 +10141,8 @@ PyUnicode_InternFromString(const char *cp) return s; } -void _Py_ReleaseInternedUnicodeStrings(void) +void +_Py_ReleaseInternedUnicodeStrings(void) { PyObject *keys; PyUnicodeObject *s;