| /*********************************************************** |
| Copyright 1991-1995 by Stichting Mathematisch Centrum, Amsterdam, |
| The Netherlands. |
| |
| All Rights Reserved |
| |
| Permission to use, copy, modify, and distribute this software and its |
| documentation for any purpose and without fee is hereby granted, |
| provided that the above copyright notice appear in all copies and that |
| both that copyright notice and this permission notice appear in |
| supporting documentation, and that the names of Stichting Mathematisch |
| Centrum or CWI or Corporation for National Research Initiatives or |
| CNRI not be used in advertising or publicity pertaining to |
| distribution of the software without specific, written prior |
| permission. |
| |
| While CWI is the initial source for this software, a modified version |
| is made available by the Corporation for National Research Initiatives |
| (CNRI) at the Internet address ftp://ftp.python.org. |
| |
| STICHTING MATHEMATISCH CENTRUM AND CNRI DISCLAIM ALL WARRANTIES WITH |
| REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF |
| MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL STICHTING MATHEMATISCH |
| CENTRUM OR CNRI BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL |
| DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR |
| PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER |
| TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR |
| PERFORMANCE OF THIS SOFTWARE. |
| |
| ******************************************************************/ |
| |
| /* strop module */ |
| |
| static char strop_module__doc__[] = |
| "Common string manipulations, optimized for speed.\n\ |
| \n\ |
| Always use \"import string\" rather than referencing\n\ |
| this module directly."; |
| |
| #include "Python.h" |
| |
| #ifdef HAVE_LIMITS_H |
| #include <limits.h> |
| #else |
| #define INT_MAX 2147483647 |
| #endif |
| |
| #include <ctype.h> |
| /* XXX This file assumes that the <ctype.h> is*() functions |
| XXX are defined for all 8-bit characters! */ |
| |
| /* The lstrip(), rstrip() and strip() functions are implemented |
| in do_strip(), which uses an additional parameter to indicate what |
| type of strip should occur. */ |
| |
| #define LEFTSTRIP 0 |
| #define RIGHTSTRIP 1 |
| #define BOTHSTRIP 2 |
| |
| |
| static PyObject * |
| split_whitespace(s, len, maxsplit) |
| char *s; |
| int len; |
| int maxsplit; |
| { |
| int i = 0, j, err; |
| int countsplit = 0; |
| PyObject* item; |
| PyObject *list = PyList_New(0); |
| |
| if (list == NULL) |
| return NULL; |
| |
| while (i < len) { |
| while (i < len && isspace(Py_CHARMASK(s[i]))) { |
| i = i+1; |
| } |
| j = i; |
| while (i < len && !isspace(Py_CHARMASK(s[i]))) { |
| i = i+1; |
| } |
| if (j < i) { |
| item = PyString_FromStringAndSize(s+j, (int)(i-j)); |
| if (item == NULL) |
| goto finally; |
| |
| err = PyList_Append(list, item); |
| Py_DECREF(item); |
| if (err < 0) |
| goto finally; |
| |
| countsplit++; |
| while (i < len && isspace(Py_CHARMASK(s[i]))) { |
| i = i+1; |
| } |
| if (maxsplit && (countsplit >= maxsplit) && i < len) { |
| item = PyString_FromStringAndSize( |
| s+i, (int)(len - i)); |
| if (item == NULL) |
| goto finally; |
| |
| err = PyList_Append(list, item); |
| Py_DECREF(item); |
| if (err < 0) |
| goto finally; |
| |
| i = len; |
| } |
| } |
| } |
| return list; |
| finally: |
| Py_DECREF(list); |
| return NULL; |
| } |
| |
| |
| static char splitfields__doc__[] = |
| "split(str [,sep [,maxsplit]]) -> list of strings\n\ |
| splitfields(str [,sep [,maxsplit]]) -> list of strings\n\ |
| \n\ |
| Return a list of the words in the string s, using sep as the\n\ |
| delimiter string. If maxsplit is nonzero, splits into at most\n\ |
| maxsplit words If sep is not specified, any whitespace string\n\ |
| is a separator. Maxsplit defaults to 0.\n\ |
| \n\ |
| (split and splitfields are synonymous)"; |
| |
| static PyObject * |
| strop_splitfields(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| int len, n, i, j, err; |
| int splitcount, maxsplit; |
| char *s, *sub; |
| PyObject *list, *item; |
| |
| sub = NULL; |
| n = 0; |
| splitcount = 0; |
| maxsplit = 0; |
| if (!PyArg_ParseTuple(args, "s#|z#i", &s, &len, &sub, &n, &maxsplit)) |
| return NULL; |
| if (sub == NULL) |
| return split_whitespace(s, len, maxsplit); |
| if (n == 0) { |
| PyErr_SetString(PyExc_ValueError, "empty separator"); |
| return NULL; |
| } |
| |
| list = PyList_New(0); |
| if (list == NULL) |
| return NULL; |
| |
| i = j = 0; |
| while (i+n <= len) { |
| if (s[i] == sub[0] && (n == 1 || memcmp(s+i, sub, n) == 0)) { |
| item = PyString_FromStringAndSize(s+j, (int)(i-j)); |
| if (item == NULL) |
| goto fail; |
| err = PyList_Append(list, item); |
| Py_DECREF(item); |
| if (err < 0) |
| goto fail; |
| i = j = i + n; |
| splitcount++; |
| if (maxsplit && (splitcount >= maxsplit)) |
| break; |
| } |
| else |
| i++; |
| } |
| item = PyString_FromStringAndSize(s+j, (int)(len-j)); |
| if (item == NULL) |
| goto fail; |
| err = PyList_Append(list, item); |
| Py_DECREF(item); |
| if (err < 0) |
| goto fail; |
| |
| return list; |
| |
| fail: |
| Py_DECREF(list); |
| return NULL; |
| } |
| |
| |
| static char joinfields__doc__[] = |
| "join(list [,sep]) -> string\n\ |
| joinfields(list [,sep]) -> string\n\ |
| \n\ |
| Return a string composed of the words in list, with\n\ |
| intervening occurences of sep. Sep defaults to a single\n\ |
| space.\n\ |
| \n\ |
| (join and joinfields are synonymous)"; |
| |
| static PyObject * |
| strop_joinfields(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| PyObject *seq; |
| char *sep = NULL; |
| int seqlen, seplen = 0; |
| int i, reslen = 0, slen = 0, sz = 100; |
| PyObject *res = NULL; |
| char* p = NULL; |
| intargfunc getitemfunc; |
| |
| if (!PyArg_ParseTuple(args, "O|s#", &seq, &sep, &seplen)) |
| return NULL; |
| if (sep == NULL) { |
| sep = " "; |
| seplen = 1; |
| } |
| |
| seqlen = PySequence_Length(seq); |
| if (seqlen < 0 && PyErr_Occurred()) |
| return NULL; |
| |
| if (seqlen == 1) { |
| /* Optimization if there's only one item */ |
| PyObject *item = PySequence_GetItem(seq, 0); |
| if (item && !PyString_Check(item)) { |
| PyErr_SetString(PyExc_TypeError, |
| "first argument must be sequence of strings"); |
| return NULL; |
| } |
| return item; |
| } |
| |
| if (!(res = PyString_FromStringAndSize((char*)NULL, sz))) |
| return NULL; |
| p = PyString_AsString(res); |
| |
| /* optimize for lists, since it's the most common case. all others |
| * (tuples and arbitrary sequences) just use the sequence abstract |
| * interface. |
| */ |
| if (PyList_Check(seq)) { |
| for (i = 0; i < seqlen; i++) { |
| PyObject *item = PyList_GET_ITEM(seq, i); |
| if (!PyString_Check(item)) { |
| PyErr_SetString(PyExc_TypeError, |
| "first argument must be sequence of strings"); |
| Py_DECREF(res); |
| return NULL; |
| } |
| slen = PyString_GET_SIZE(item); |
| while (reslen + slen + seplen >= sz) { |
| if (_PyString_Resize(&res, sz * 2)) { |
| Py_DECREF(res); |
| return NULL; |
| } |
| sz *= 2; |
| p = PyString_AsString(res) + reslen; |
| } |
| if (i > 0) { |
| memcpy(p, sep, seplen); |
| p += seplen; |
| reslen += seplen; |
| } |
| memcpy(p, PyString_AS_STRING(item), slen); |
| p += slen; |
| reslen += slen; |
| } |
| if (_PyString_Resize(&res, reslen)) { |
| Py_DECREF(res); |
| res = NULL; |
| } |
| return res; |
| } |
| else if (!PySequence_Check(seq)) { |
| PyErr_SetString(PyExc_TypeError, |
| "first argument must be a sequence"); |
| return NULL; |
| } |
| /* type safe */ |
| getitemfunc = seq->ob_type->tp_as_sequence->sq_item; |
| for (i = 0; i < seqlen; i++) { |
| PyObject *item = getitemfunc(seq, i); |
| if (!item || !PyString_Check(item)) { |
| PyErr_SetString(PyExc_TypeError, |
| "first argument must be sequence of strings"); |
| Py_DECREF(res); |
| Py_XDECREF(item); |
| return NULL; |
| } |
| slen = PyString_GET_SIZE(item); |
| while (reslen + slen + seplen >= sz) { |
| if (_PyString_Resize(&res, sz * 2)) { |
| Py_DECREF(res); |
| Py_DECREF(item); |
| return NULL; |
| } |
| sz *= 2; |
| p = PyString_AsString(res) + reslen; |
| } |
| if (i > 0) { |
| memcpy(p, sep, seplen); |
| p += seplen; |
| reslen += seplen; |
| } |
| memcpy(p, PyString_AS_STRING(item), slen); |
| p += slen; |
| reslen += slen; |
| Py_DECREF(item); |
| } |
| if (_PyString_Resize(&res, reslen)) { |
| Py_DECREF(res); |
| res = NULL; |
| } |
| return res; |
| } |
| |
| |
| static char find__doc__[] = |
| "find(s, sub [,start [,end]]) -> in\n\ |
| \n\ |
| Return the lowest index in s where substring sub is found,\n\ |
| such that sub is contained within s[start,end]. Optional\n\ |
| arguments start and end are interpreted as in slice notation.\n\ |
| \n\ |
| Return -1 on failure."; |
| |
| static PyObject * |
| strop_find(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| char *s, *sub; |
| int len, n, i = 0, last = INT_MAX; |
| |
| if (!PyArg_ParseTuple(args, "s#s#|ii", &s, &len, &sub, &n, &i, &last)) |
| return NULL; |
| |
| if (last > len) |
| last = len; |
| if (last < 0) |
| last += len; |
| if (last < 0) |
| last = 0; |
| if (i < 0) |
| i += len; |
| if (i < 0) |
| i = 0; |
| |
| if (n == 0 && i <= last) |
| return PyInt_FromLong((long)i); |
| |
| last -= n; |
| for (; i <= last; ++i) |
| if (s[i] == sub[0] && |
| (n == 1 || memcmp(&s[i+1], &sub[1], n-1) == 0)) |
| return PyInt_FromLong((long)i); |
| |
| return PyInt_FromLong(-1L); |
| } |
| |
| |
| static char rfind__doc__[] = |
| "rfind(s, sub [,start [,end]]) -> int\n\ |
| \n\ |
| Return the highest index in s where substring sub is found,\n\ |
| such that sub is contained within s[start,end]. Optional\n\ |
| arguments start and end are interpreted as in slice notation.\n\ |
| \n\ |
| Return -1 on failure."; |
| |
| static PyObject * |
| strop_rfind(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| char *s, *sub; |
| int len, n, j; |
| int i = 0, last = INT_MAX; |
| |
| if (!PyArg_ParseTuple(args, "s#s#|ii", &s, &len, &sub, &n, &i, &last)) |
| return NULL; |
| |
| if (last > len) |
| last = len; |
| if (last < 0) |
| last += len; |
| if (last < 0) |
| last = 0; |
| if (i < 0) |
| i += len; |
| if (i < 0) |
| i = 0; |
| |
| if (n == 0 && i <= last) |
| return PyInt_FromLong((long)last); |
| |
| for (j = last-n; j >= i; --j) |
| if (s[j] == sub[0] && |
| (n == 1 || memcmp(&s[j+1], &sub[1], n-1) == 0)) |
| return PyInt_FromLong((long)j); |
| |
| return PyInt_FromLong(-1L); |
| } |
| |
| |
| static PyObject * |
| do_strip(args, striptype) |
| PyObject *args; |
| int striptype; |
| { |
| char *s; |
| int len, i, j; |
| |
| |
| if (!PyArg_Parse(args, "s#", &s, &len)) |
| return NULL; |
| |
| i = 0; |
| if (striptype != RIGHTSTRIP) { |
| while (i < len && isspace(Py_CHARMASK(s[i]))) { |
| i++; |
| } |
| } |
| |
| j = len; |
| if (striptype != LEFTSTRIP) { |
| do { |
| j--; |
| } while (j >= i && isspace(Py_CHARMASK(s[j]))); |
| j++; |
| } |
| |
| if (i == 0 && j == len) { |
| Py_INCREF(args); |
| return args; |
| } |
| else |
| return PyString_FromStringAndSize(s+i, j-i); |
| } |
| |
| |
| static char strip__doc__[] = |
| "strip(s) -> string\n\ |
| \n\ |
| Return a copy of the string s with leading and trailing\n\ |
| whitespace removed."; |
| |
| static PyObject * |
| strop_strip(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| return do_strip(args, BOTHSTRIP); |
| } |
| |
| |
| static char lstrip__doc__[] = |
| "lstrip(s) -> string\n\ |
| \n\ |
| Return a copy of the string s with leading whitespace removed."; |
| |
| static PyObject * |
| strop_lstrip(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| return do_strip(args, LEFTSTRIP); |
| } |
| |
| |
| static char rstrip__doc__[] = |
| "rstrip(s) -> string\n\ |
| \n\ |
| Return a copy of the string s with trailing whitespace removed."; |
| |
| static PyObject * |
| strop_rstrip(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| return do_strip(args, RIGHTSTRIP); |
| } |
| |
| |
| static char lower__doc__[] = |
| "lower(s) -> string\n\ |
| \n\ |
| Return a copy of the string s converted to lowercase."; |
| |
| static PyObject * |
| strop_lower(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| char *s, *s_new; |
| int i, n; |
| PyObject *new; |
| int changed; |
| |
| if (!PyArg_Parse(args, "s#", &s, &n)) |
| return NULL; |
| new = PyString_FromStringAndSize(NULL, n); |
| if (new == NULL) |
| return NULL; |
| s_new = PyString_AsString(new); |
| changed = 0; |
| for (i = 0; i < n; i++) { |
| int c = Py_CHARMASK(*s++); |
| if (isupper(c)) { |
| changed = 1; |
| *s_new = tolower(c); |
| } else |
| *s_new = c; |
| s_new++; |
| } |
| if (!changed) { |
| Py_DECREF(new); |
| Py_INCREF(args); |
| return args; |
| } |
| return new; |
| } |
| |
| |
| static char upper__doc__[] = |
| "upper(s) -> string\n\ |
| \n\ |
| Return a copy of the string s converted to uppercase."; |
| |
| static PyObject * |
| strop_upper(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| char *s, *s_new; |
| int i, n; |
| PyObject *new; |
| int changed; |
| |
| if (!PyArg_Parse(args, "s#", &s, &n)) |
| return NULL; |
| new = PyString_FromStringAndSize(NULL, n); |
| if (new == NULL) |
| return NULL; |
| s_new = PyString_AsString(new); |
| changed = 0; |
| for (i = 0; i < n; i++) { |
| int c = Py_CHARMASK(*s++); |
| if (islower(c)) { |
| changed = 1; |
| *s_new = toupper(c); |
| } else |
| *s_new = c; |
| s_new++; |
| } |
| if (!changed) { |
| Py_DECREF(new); |
| Py_INCREF(args); |
| return args; |
| } |
| return new; |
| } |
| |
| |
| static char capitalize__doc__[] = |
| "capitalize(s) -> string\n\ |
| \n\ |
| Return a copy of the string s with only its first character\n\ |
| capitalized."; |
| |
| static PyObject * |
| strop_capitalize(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| char *s, *s_new; |
| int i, n; |
| PyObject *new; |
| int changed; |
| |
| if (!PyArg_Parse(args, "s#", &s, &n)) |
| return NULL; |
| new = PyString_FromStringAndSize(NULL, n); |
| if (new == NULL) |
| return NULL; |
| s_new = PyString_AsString(new); |
| changed = 0; |
| if (0 < n) { |
| int c = Py_CHARMASK(*s++); |
| if (islower(c)) { |
| changed = 1; |
| *s_new = toupper(c); |
| } else |
| *s_new = c; |
| s_new++; |
| } |
| for (i = 1; i < n; i++) { |
| int c = Py_CHARMASK(*s++); |
| if (isupper(c)) { |
| changed = 1; |
| *s_new = tolower(c); |
| } else |
| *s_new = c; |
| s_new++; |
| } |
| if (!changed) { |
| Py_DECREF(new); |
| Py_INCREF(args); |
| return args; |
| } |
| return new; |
| } |
| |
| |
| static char swapcase__doc__[] = |
| "swapcase(s) -> string\n\ |
| \n\ |
| Return a copy of the string s with upper case characters\n\ |
| converted to lowercase and vice versa."; |
| |
| static PyObject * |
| strop_swapcase(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| char *s, *s_new; |
| int i, n; |
| PyObject *new; |
| int changed; |
| |
| if (!PyArg_Parse(args, "s#", &s, &n)) |
| return NULL; |
| new = PyString_FromStringAndSize(NULL, n); |
| if (new == NULL) |
| return NULL; |
| s_new = PyString_AsString(new); |
| changed = 0; |
| for (i = 0; i < n; i++) { |
| int c = Py_CHARMASK(*s++); |
| if (islower(c)) { |
| changed = 1; |
| *s_new = toupper(c); |
| } |
| else if (isupper(c)) { |
| changed = 1; |
| *s_new = tolower(c); |
| } |
| else |
| *s_new = c; |
| s_new++; |
| } |
| if (!changed) { |
| Py_DECREF(new); |
| Py_INCREF(args); |
| return args; |
| } |
| return new; |
| } |
| |
| |
| static char atoi__doc__[] = |
| "atoi(s [,base]) -> int\n\ |
| \n\ |
| Return the integer represented by the string s in the given\n\ |
| base, which defaults to 10. The string s must consist of one\n\ |
| or more digits, possibly preceded by a sign. If base is 0, it\n\ |
| is chosen from the leading characters of s, 0 for octal, 0x or\n\ |
| 0X for hexadecimal. If base is 16, a preceding 0x or 0X is\n\ |
| accepted."; |
| |
| static PyObject * |
| strop_atoi(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| extern long PyOS_strtol Py_PROTO((const char *, char **, int)); |
| extern unsigned long |
| PyOS_strtoul Py_PROTO((const char *, char **, int)); |
| char *s, *end; |
| int base = 10; |
| long x; |
| char buffer[256]; /* For errors */ |
| |
| if (!PyArg_ParseTuple(args, "s|i", &s, &base)) |
| return NULL; |
| |
| if ((base != 0 && base < 2) || base > 36) { |
| PyErr_SetString(PyExc_ValueError, "invalid base for atoi()"); |
| return NULL; |
| } |
| |
| while (*s && isspace(Py_CHARMASK(*s))) |
| s++; |
| if (s[0] == '\0') { |
| PyErr_SetString(PyExc_ValueError, "empty string for atoi()"); |
| return NULL; |
| } |
| errno = 0; |
| if (base == 0 && s[0] == '0') |
| x = (long) PyOS_strtoul(s, &end, base); |
| else |
| x = PyOS_strtol(s, &end, base); |
| while (*end && isspace(Py_CHARMASK(*end))) |
| end++; |
| if (*end != '\0') { |
| sprintf(buffer, "invalid literal for atoi(): %.200s", s); |
| PyErr_SetString(PyExc_ValueError, buffer); |
| return NULL; |
| } |
| else if (errno != 0) { |
| sprintf(buffer, "atoi() literal too large: %.200s", s); |
| PyErr_SetString(PyExc_ValueError, buffer); |
| return NULL; |
| } |
| return PyInt_FromLong(x); |
| } |
| |
| |
| static char atol__doc__[] = |
| "atol(s [,base]) -> long\n\ |
| \n\ |
| Return the long integer represented by the string s in the\n\ |
| given base, which defaults to 10. The string s must consist\n\ |
| of one or more digits, possibly preceded by a sign. If base\n\ |
| is 0, it is chosen from the leading characters of s, 0 for\n\ |
| octal, 0x or 0X for hexadecimal. If base is 16, a preceding\n\ |
| 0x or 0X is accepted. A trailing L or l is not accepted,\n\ |
| unless base is 0."; |
| |
| static PyObject * |
| strop_atol(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| char *s, *end; |
| int base = 10; |
| PyObject *x; |
| char buffer[256]; /* For errors */ |
| |
| if (!PyArg_ParseTuple(args, "s|i", &s, &base)) |
| return NULL; |
| |
| if ((base != 0 && base < 2) || base > 36) { |
| PyErr_SetString(PyExc_ValueError, "invalid base for atol()"); |
| return NULL; |
| } |
| |
| while (*s && isspace(Py_CHARMASK(*s))) |
| s++; |
| if (s[0] == '\0') { |
| PyErr_SetString(PyExc_ValueError, "empty string for atol()"); |
| return NULL; |
| } |
| x = PyLong_FromString(s, &end, base); |
| if (x == NULL) |
| return NULL; |
| if (base == 0 && (*end == 'l' || *end == 'L')) |
| end++; |
| while (*end && isspace(Py_CHARMASK(*end))) |
| end++; |
| if (*end != '\0') { |
| sprintf(buffer, "invalid literal for atol(): %.200s", s); |
| PyErr_SetString(PyExc_ValueError, buffer); |
| Py_DECREF(x); |
| return NULL; |
| } |
| return x; |
| } |
| |
| |
| static char atof__doc__[] = |
| "atof(s) -> float\n\ |
| \n\ |
| Return the floating point number represented by the string s."; |
| |
| static PyObject * |
| strop_atof(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| extern double strtod Py_PROTO((const char *, char **)); |
| char *s, *end; |
| double x; |
| char buffer[256]; /* For errors */ |
| |
| if (!PyArg_ParseTuple(args, "s", &s)) |
| return NULL; |
| while (*s && isspace(Py_CHARMASK(*s))) |
| s++; |
| if (s[0] == '\0') { |
| PyErr_SetString(PyExc_ValueError, "empty string for atof()"); |
| return NULL; |
| } |
| errno = 0; |
| PyFPE_START_PROTECT("strop_atof", return 0) |
| x = strtod(s, &end); |
| PyFPE_END_PROTECT(x) |
| while (*end && isspace(Py_CHARMASK(*end))) |
| end++; |
| if (*end != '\0') { |
| sprintf(buffer, "invalid literal for atof(): %.200s", s); |
| PyErr_SetString(PyExc_ValueError, buffer); |
| return NULL; |
| } |
| else if (errno != 0) { |
| sprintf(buffer, "atof() literal too large: %.200s", s); |
| PyErr_SetString(PyExc_ValueError, buffer); |
| return NULL; |
| } |
| return PyFloat_FromDouble(x); |
| } |
| |
| |
| static char maketrans__doc__[] = |
| "maketrans(frm, to) -> string\n\ |
| \n\ |
| Return a translation table (a string of 256 bytes long)\n\ |
| suitable for use in string.translate. The strings frm and to\n\ |
| must be of the same length."; |
| |
| static PyObject * |
| strop_maketrans(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| unsigned char *c, *from=NULL, *to=NULL; |
| int i, fromlen=0, tolen=0; |
| PyObject *result; |
| |
| if (!PyArg_ParseTuple(args, "s#s#", &from, &fromlen, &to, &tolen)) |
| return NULL; |
| |
| if (fromlen != tolen) { |
| PyErr_SetString(PyExc_ValueError, |
| "maketrans arguments must have same length"); |
| return NULL; |
| } |
| |
| result = PyString_FromStringAndSize((char *)NULL, 256); |
| if (result == NULL) |
| return NULL; |
| c = (unsigned char *) PyString_AS_STRING((PyStringObject *)result); |
| for (i = 0; i < 256; i++) |
| c[i]=(unsigned char)i; |
| for (i = 0; i < fromlen; i++) |
| c[from[i]]=to[i]; |
| |
| return result; |
| } |
| |
| |
| static char translate__doc__[] = |
| "translate(s,table [,deletechars]) -> string\n\ |
| \n\ |
| Return a copy of the string s, where all characters occurring\n\ |
| in the optional argument deletechars are removed, and the\n\ |
| remaining characters have been mapped through the given\n\ |
| translation table, which must be a string of length 256."; |
| |
| static PyObject * |
| strop_translate(self, args) |
| PyObject *self; |
| PyObject *args; |
| { |
| register char *input, *table, *output; |
| register int i, c, changed = 0; |
| PyObject *input_obj; |
| char *table1, *output_start, *del_table=NULL; |
| int inlen, tablen, dellen = 0; |
| PyObject *result; |
| int trans_table[256]; |
| |
| if (!PyArg_ParseTuple(args, "Ss#|s#", &input_obj, |
| &table1, &tablen, &del_table, &dellen)) |
| return NULL; |
| if (tablen != 256) { |
| PyErr_SetString(PyExc_ValueError, |
| "translation table must be 256 characters long"); |
| return NULL; |
| } |
| |
| table = table1; |
| inlen = PyString_Size(input_obj); |
| result = PyString_FromStringAndSize((char *)NULL, inlen); |
| if (result == NULL) |
| return NULL; |
| output_start = output = PyString_AsString(result); |
| input = PyString_AsString(input_obj); |
| |
| if (dellen == 0) { |
| /* If no deletions are required, use faster code */ |
| for (i = inlen; --i >= 0; ) { |
| c = Py_CHARMASK(*input++); |
| if (Py_CHARMASK((*output++ = table[c])) != c) |
| changed = 1; |
| } |
| if (changed) |
| return result; |
| Py_DECREF(result); |
| Py_INCREF(input_obj); |
| return input_obj; |
| } |
| |
| for (i = 0; i < 256; i++) |
| trans_table[i] = Py_CHARMASK(table[i]); |
| |
| for (i = 0; i < dellen; i++) |
| trans_table[(int) Py_CHARMASK(del_table[i])] = -1; |
| |
| for (i = inlen; --i >= 0; ) { |
| c = Py_CHARMASK(*input++); |
| if (trans_table[c] != -1) |
| if (Py_CHARMASK(*output++ = (char)trans_table[c]) == c) |
| continue; |
| changed = 1; |
| } |
| if (!changed) { |
| Py_DECREF(result); |
| Py_INCREF(input_obj); |
| return input_obj; |
| } |
| /* Fix the size of the resulting string */ |
| if (inlen > 0 &&_PyString_Resize(&result, output-output_start)) |
| return NULL; |
| return result; |
| } |
| |
| |
| /* What follows is used for implementing replace(). Perry Stoll. */ |
| |
| /* |
| mymemfind |
| |
| strstr replacement for arbitrary blocks of memory. |
| |
| Locates the first occurance in the memory pointed to by MEM of the |
| contents of memory pointed to by PAT. Returns the index into MEM if |
| found, or -1 if not found. If len of PAT is greater than length of |
| MEM, the function returns -1. |
| */ |
| static int mymemfind(mem, len, pat, pat_len) |
| char *mem; |
| int len; |
| char *pat; |
| int pat_len; |
| { |
| register int ii; |
| |
| /* pattern can not occur in the last pat_len-1 chars */ |
| len -= pat_len; |
| |
| for (ii = 0; ii <= len; ii++) { |
| if (mem[ii] == pat[0] && |
| (pat_len == 1 || |
| memcmp(&mem[ii+1], &pat[1], pat_len-1) == 0)) { |
| return ii; |
| } |
| } |
| return -1; |
| } |
| |
| /* |
| mymemcnt |
| |
| Return the number of distinct times PAT is found in MEM. |
| meaning mem=1111 and pat==11 returns 2. |
| mem=11111 and pat==11 also return 2. |
| */ |
| static int mymemcnt(mem, len, pat, pat_len) |
| char *mem; |
| int len; |
| char *pat; |
| int pat_len; |
| { |
| register int offset = 0; |
| int nfound = 0; |
| |
| while (len >= 0) { |
| offset = mymemfind(mem, len, pat, pat_len); |
| if (offset == -1) |
| break; |
| mem += offset + pat_len; |
| len -= offset + pat_len; |
| nfound++; |
| } |
| return nfound; |
| } |
| |
| /* |
| mymemreplace |
| |
| Return a string in which all occurences of PAT in memory STR are |
| replaced with SUB. |
| |
| If length of PAT is less than length of STR or there are no occurences |
| of PAT in STR, then the original string is returned. Otherwise, a new |
| string is allocated here and returned. |
| |
| on return, out_len is: |
| the length of output string, or |
| -1 if the input string is returned, or |
| unchanged if an error occurs (no memory). |
| |
| return value is: |
| the new string allocated locally, or |
| NULL if an error occurred. |
| */ |
| static char *mymemreplace(str, len, pat, pat_len, sub, sub_len, count, out_len) |
| char *str; |
| int len; /* input string */ |
| char *pat; |
| int pat_len; /* pattern string to find */ |
| char *sub; |
| int sub_len; /* substitution string */ |
| int count; /* number of replacements, 0 == all */ |
| int *out_len; |
| |
| { |
| char *out_s; |
| char *new_s; |
| int nfound, offset, new_len; |
| |
| if (len == 0 || pat_len > len) |
| goto return_same; |
| |
| /* find length of output string */ |
| nfound = mymemcnt(str, len, pat, pat_len); |
| if (count > 0) |
| nfound = nfound > count ? count : nfound; |
| if (nfound == 0) |
| goto return_same; |
| new_len = len + nfound*(sub_len - pat_len); |
| |
| new_s = (char *)malloc(new_len); |
| if (new_s == NULL) return NULL; |
| |
| *out_len = new_len; |
| out_s = new_s; |
| |
| while (len > 0) { |
| /* find index of next instance of pattern */ |
| offset = mymemfind(str, len, pat, pat_len); |
| /* if not found, break out of loop */ |
| if (offset == -1) break; |
| |
| /* copy non matching part of input string */ |
| memcpy(new_s, str, offset); /* copy part of str before pat */ |
| str += offset + pat_len; /* move str past pattern */ |
| len -= offset + pat_len; /* reduce length of str remaining */ |
| |
| /* copy substitute into the output string */ |
| new_s += offset; /* move new_s to dest for sub string */ |
| memcpy(new_s, sub, sub_len); /* copy substring into new_s */ |
| new_s += sub_len; /* offset new_s past sub string */ |
| |
| /* break when we've done count replacements */ |
| if (--count == 0) break; |
| } |
| /* copy any remaining values into output string */ |
| if (len > 0) |
| memcpy(new_s, str, len); |
| return out_s; |
| |
| return_same: |
| *out_len = -1; |
| return str; |
| } |
| |
| |
| static char replace__doc__[] = |
| "replace (str, old, new[, maxsplit]) -> string\n\ |
| \n\ |
| Return a copy of string str with all occurrences of substring\n\ |
| old replaced by new. If the optional argument maxsplit is\n\ |
| given, only the first maxsplit occurrences are replaced."; |
| |
| static PyObject * |
| strop_replace(self, args) |
| PyObject *self; /* Not used */ |
| PyObject *args; |
| { |
| char *str, *pat,*sub,*new_s; |
| int len,pat_len,sub_len,out_len; |
| int count = 0; |
| PyObject *new; |
| |
| if (!PyArg_ParseTuple(args, "s#s#s#|i", |
| &str, &len, &pat, &pat_len, &sub, &sub_len, |
| &count)) |
| return NULL; |
| new_s = mymemreplace(str,len,pat,pat_len,sub,sub_len,count,&out_len); |
| if (new_s == NULL) { |
| PyErr_NoMemory(); |
| return NULL; |
| } |
| if (out_len == -1) { |
| /* we're returning another reference to the input string */ |
| new = PyTuple_GetItem(args, 0); |
| Py_XINCREF(new); |
| } |
| else { |
| new = PyString_FromStringAndSize(new_s, out_len); |
| free(new_s); |
| } |
| return new; |
| } |
| |
| |
| /* List of functions defined in the module */ |
| |
| static PyMethodDef |
| strop_methods[] = { |
| {"atof", strop_atof, 1, atof__doc__}, |
| {"atoi", strop_atoi, 1, atoi__doc__}, |
| {"atol", strop_atol, 1, atol__doc__}, |
| {"capitalize", strop_capitalize, 0, capitalize__doc__}, |
| {"find", strop_find, 1, find__doc__}, |
| {"join", strop_joinfields, 1, joinfields__doc__}, |
| {"joinfields", strop_joinfields, 1, joinfields__doc__}, |
| {"lstrip", strop_lstrip, 0, lstrip__doc__}, |
| {"lower", strop_lower, 0, lower__doc__}, |
| {"maketrans", strop_maketrans, 1, maketrans__doc__}, |
| {"replace", strop_replace, 1, replace__doc__}, |
| {"rfind", strop_rfind, 1, rfind__doc__}, |
| {"rstrip", strop_rstrip, 0, rstrip__doc__}, |
| {"split", strop_splitfields, 1, splitfields__doc__}, |
| {"splitfields", strop_splitfields, 1, splitfields__doc__}, |
| {"strip", strop_strip, 0, strip__doc__}, |
| {"swapcase", strop_swapcase, 0, swapcase__doc__}, |
| {"translate", strop_translate, 1, translate__doc__}, |
| {"upper", strop_upper, 0, upper__doc__}, |
| {NULL, NULL} /* sentinel */ |
| }; |
| |
| |
| void |
| initstrop() |
| { |
| PyObject *m, *d, *s; |
| char buf[256]; |
| int c, n; |
| m = Py_InitModule4("strop", strop_methods, strop_module__doc__, |
| (PyObject*)NULL, PYTHON_API_VERSION); |
| d = PyModule_GetDict(m); |
| |
| /* Create 'whitespace' object */ |
| n = 0; |
| for (c = 0; c < 256; c++) { |
| if (isspace(c)) |
| buf[n++] = c; |
| } |
| s = PyString_FromStringAndSize(buf, n); |
| if (s) { |
| PyDict_SetItemString(d, "whitespace", s); |
| Py_DECREF(s); |
| } |
| /* Create 'lowercase' object */ |
| n = 0; |
| for (c = 0; c < 256; c++) { |
| if (islower(c)) |
| buf[n++] = c; |
| } |
| s = PyString_FromStringAndSize(buf, n); |
| if (s) { |
| PyDict_SetItemString(d, "lowercase", s); |
| Py_DECREF(s); |
| } |
| |
| /* Create 'uppercase' object */ |
| n = 0; |
| for (c = 0; c < 256; c++) { |
| if (isupper(c)) |
| buf[n++] = c; |
| } |
| s = PyString_FromStringAndSize(buf, n); |
| if (s) { |
| PyDict_SetItemString(d, "uppercase", s); |
| Py_DECREF(s); |
| } |
| |
| if (PyErr_Occurred()) |
| Py_FatalError("can't initialize module strop"); |
| } |