| 1 | // libc.cc: Replacement for pyext/libc.c
|
| 2 |
|
| 3 | #include "cpp/libc.h"
|
| 4 |
|
| 5 | #include <errno.h>
|
| 6 | #include <fnmatch.h>
|
| 7 | #include <glob.h>
|
| 8 | #include <locale.h>
|
| 9 | #include <regex.h>
|
| 10 | #include <sys/ioctl.h>
|
| 11 | #include <unistd.h> // gethostname()
|
| 12 | #include <wchar.h>
|
| 13 |
|
| 14 | namespace libc {
|
| 15 |
|
| 16 | BigStr* gethostname() {
|
| 17 | // Note: Fixed issue #1656 - OS X and FreeBSD don't have HOST_NAME_MAX
|
| 18 | // https://reviews.freebsd.org/D30062
|
| 19 | BigStr* result = OverAllocatedStr(_POSIX_HOST_NAME_MAX);
|
| 20 | int status = ::gethostname(result->data_, _POSIX_HOST_NAME_MAX);
|
| 21 | if (status != 0) {
|
| 22 | throw Alloc<OSError>(errno);
|
| 23 | }
|
| 24 | // Important: set the length of the string!
|
| 25 | result->MaybeShrink(strlen(result->data_));
|
| 26 | return result;
|
| 27 | }
|
| 28 |
|
| 29 | BigStr* realpath(BigStr* path) {
|
| 30 | BigStr* result = OverAllocatedStr(PATH_MAX);
|
| 31 | char* p = ::realpath(path->data_, result->data_);
|
| 32 | if (p == nullptr) {
|
| 33 | throw Alloc<OSError>(errno);
|
| 34 | }
|
| 35 | result->MaybeShrink(strlen(result->data_));
|
| 36 | return result;
|
| 37 | }
|
| 38 |
|
| 39 | int fnmatch(BigStr* pat, BigStr* str, int flags) {
|
| 40 | #ifdef FNM_EXTMATCH
|
| 41 | flags |= FNM_EXTMATCH;
|
| 42 | #else
|
| 43 | // TODO: We should detect this at ./configure time, and then maybe flag these
|
| 44 | // at parse time, not runtime
|
| 45 | #endif
|
| 46 |
|
| 47 | int result = ::fnmatch(pat->data_, str->data_, flags);
|
| 48 | switch (result) {
|
| 49 | case 0:
|
| 50 | return 1;
|
| 51 | case FNM_NOMATCH:
|
| 52 | return 0;
|
| 53 | default:
|
| 54 | // Other error
|
| 55 | return -1;
|
| 56 | }
|
| 57 | }
|
| 58 |
|
| 59 | List<BigStr*>* glob(BigStr* pat) {
|
| 60 | glob_t results;
|
| 61 | // Hm, it's weird that the first one can't be called with GLOB_APPEND. You
|
| 62 | // get a segfault.
|
| 63 | int flags = 0;
|
| 64 | // int flags = GLOB_APPEND;
|
| 65 | // flags |= GLOB_NOMAGIC;
|
| 66 | int ret = glob(pat->data_, flags, NULL, &results);
|
| 67 |
|
| 68 | const char* err_str = NULL;
|
| 69 | switch (ret) {
|
| 70 | case 0: // no error
|
| 71 | break;
|
| 72 | case GLOB_ABORTED:
|
| 73 | err_str = "read error";
|
| 74 | break;
|
| 75 | case GLOB_NOMATCH:
|
| 76 | // No error, because not matching isn't necessarily a problem.
|
| 77 | // NOTE: This can be turned on to log overaggressive calls to glob().
|
| 78 | // err_str = "nothing matched";
|
| 79 | break;
|
| 80 | case GLOB_NOSPACE:
|
| 81 | err_str = "no dynamic memory";
|
| 82 | break;
|
| 83 | default:
|
| 84 | err_str = "unknown problem";
|
| 85 | break;
|
| 86 | }
|
| 87 | if (err_str) {
|
| 88 | throw Alloc<RuntimeError>(StrFromC(err_str));
|
| 89 | }
|
| 90 |
|
| 91 | // http://stackoverflow.com/questions/3512414/does-this-pylist-appendlist-py-buildvalue-leak
|
| 92 | size_t n = results.gl_pathc;
|
| 93 | auto matches = NewList<BigStr*>();
|
| 94 |
|
| 95 | // Print array of results
|
| 96 | size_t i;
|
| 97 | for (i = 0; i < n; i++) {
|
| 98 | const char* m = results.gl_pathv[i];
|
| 99 | matches->append(StrFromC(m));
|
| 100 | }
|
| 101 | globfree(&results);
|
| 102 |
|
| 103 | return matches;
|
| 104 | }
|
| 105 |
|
| 106 | // Raises RuntimeError if the pattern is invalid. TODO: Use a different
|
| 107 | // exception?
|
| 108 | List<int>* regex_search(BigStr* pattern, int cflags, BigStr* str, int eflags,
|
| 109 | int pos) {
|
| 110 | cflags |= REG_EXTENDED;
|
| 111 | regex_t pat;
|
| 112 | int status = regcomp(&pat, pattern->data_, cflags);
|
| 113 | if (status != 0) {
|
| 114 | char error_desc[50];
|
| 115 | regerror(status, &pat, error_desc, 50);
|
| 116 |
|
| 117 | char error_message[80];
|
| 118 | snprintf(error_message, 80, "Invalid regex %s (%s)", pattern->data_,
|
| 119 | error_desc);
|
| 120 |
|
| 121 | throw Alloc<ValueError>(StrFromC(error_message));
|
| 122 | }
|
| 123 | // log("pat = %d, str = %d", len(pattern), len(str));
|
| 124 |
|
| 125 | int num_groups = pat.re_nsub + 1; // number of captures
|
| 126 |
|
| 127 | List<int>* indices = NewList<int>();
|
| 128 | indices->reserve(num_groups * 2);
|
| 129 |
|
| 130 | const char* s = str->data_;
|
| 131 | regmatch_t* pmatch =
|
| 132 | static_cast<regmatch_t*>(malloc(sizeof(regmatch_t) * num_groups));
|
| 133 | bool match = regexec(&pat, s + pos, num_groups, pmatch, eflags) == 0;
|
| 134 | if (match) {
|
| 135 | int i;
|
| 136 | for (i = 0; i < num_groups; i++) {
|
| 137 | int start = pmatch[i].rm_so;
|
| 138 | if (start != -1) {
|
| 139 | start += pos;
|
| 140 | }
|
| 141 | indices->append(start);
|
| 142 |
|
| 143 | int end = pmatch[i].rm_eo;
|
| 144 | if (end != -1) {
|
| 145 | end += pos;
|
| 146 | }
|
| 147 | indices->append(end);
|
| 148 | }
|
| 149 | }
|
| 150 |
|
| 151 | free(pmatch);
|
| 152 | regfree(&pat);
|
| 153 |
|
| 154 | if (!match) {
|
| 155 | return nullptr;
|
| 156 | }
|
| 157 |
|
| 158 | return indices;
|
| 159 | }
|
| 160 |
|
| 161 | // For ${//}, the number of groups is always 1, so we want 2 match position
|
| 162 | // results -- the whole regex (which we ignore), and then first group.
|
| 163 | //
|
| 164 | // For [[ =~ ]], do we need to count how many matches the user gave?
|
| 165 |
|
| 166 | const int NMATCH = 2;
|
| 167 |
|
| 168 | // Odd: This a Tuple2* not Tuple2 because it's Optional[Tuple2]!
|
| 169 | Tuple2<int, int>* regex_first_group_match(BigStr* pattern, BigStr* str,
|
| 170 | int pos) {
|
| 171 | regex_t pat;
|
| 172 | regmatch_t m[NMATCH];
|
| 173 |
|
| 174 | // Could have been checked by regex_parse for [[ =~ ]], but not for glob
|
| 175 | // patterns like ${foo/x*/y}.
|
| 176 |
|
| 177 | if (regcomp(&pat, pattern->data_, REG_EXTENDED) != 0) {
|
| 178 | throw Alloc<RuntimeError>(
|
| 179 | StrFromC("Invalid regex syntax (func_regex_first_group_match)"));
|
| 180 | }
|
| 181 |
|
| 182 | // Match at offset 'pos'
|
| 183 | int result = regexec(&pat, str->data_ + pos, NMATCH, m, 0 /*flags*/);
|
| 184 | regfree(&pat);
|
| 185 |
|
| 186 | if (result != 0) {
|
| 187 | return nullptr;
|
| 188 | }
|
| 189 |
|
| 190 | // Assume there is a match
|
| 191 | regoff_t start = m[1].rm_so;
|
| 192 | regoff_t end = m[1].rm_eo;
|
| 193 | Tuple2<int, int>* tup = Alloc<Tuple2<int, int>>(pos + start, pos + end);
|
| 194 |
|
| 195 | return tup;
|
| 196 | }
|
| 197 |
|
| 198 | int wcswidth(BigStr* s) {
|
| 199 | // Behavior of mbstowcs() depends on LC_CTYPE
|
| 200 |
|
| 201 | // Calculate length first
|
| 202 | int num_wide_chars = ::mbstowcs(NULL, s->data_, 0);
|
| 203 | if (num_wide_chars == -1) {
|
| 204 | throw Alloc<UnicodeError>(StrFromC("mbstowcs() 1"));
|
| 205 | }
|
| 206 |
|
| 207 | // Allocate buffer
|
| 208 | int buf_size = (num_wide_chars + 1) * sizeof(wchar_t);
|
| 209 | wchar_t* wide_chars = static_cast<wchar_t*>(malloc(buf_size));
|
| 210 | DCHECK(wide_chars != nullptr);
|
| 211 |
|
| 212 | // Convert to wide chars
|
| 213 | num_wide_chars = ::mbstowcs(wide_chars, s->data_, num_wide_chars);
|
| 214 | if (num_wide_chars == -1) {
|
| 215 | free(wide_chars); // cleanup
|
| 216 |
|
| 217 | throw Alloc<UnicodeError>(StrFromC("mbstowcs() 2"));
|
| 218 | }
|
| 219 |
|
| 220 | // Find number of columns
|
| 221 | int width = ::wcswidth(wide_chars, num_wide_chars);
|
| 222 | if (width == -1) {
|
| 223 | free(wide_chars); // cleanup
|
| 224 |
|
| 225 | // unprintable chars
|
| 226 | throw Alloc<UnicodeError>(StrFromC("wcswidth()"));
|
| 227 | }
|
| 228 |
|
| 229 | free(wide_chars);
|
| 230 | return width;
|
| 231 | }
|
| 232 |
|
| 233 | int get_terminal_width() {
|
| 234 | struct winsize w;
|
| 235 | if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &w) == -1) {
|
| 236 | throw Alloc<IOError>(errno);
|
| 237 | }
|
| 238 | return w.ws_col;
|
| 239 | }
|
| 240 |
|
| 241 | } // namespace libc
|