Annotation of parser3/src/main/pa_common.C, revision 1.341
1.15 paf 1: /** @file
1.16 paf 2: Parser: commonly functions.
3:
1.334 moko 4: Copyright (c) 2000-2024 Art. Lebedev Studio (http://www.artlebedev.com)
1.327 moko 5: Authors: Konstantin Morshnev <moko@design.ru>, Alexandr Petrosian <paf@design.ru>
1.306 moko 6: */
1.210 paf 7:
1.1 paf 8: #include "pa_common.h"
1.4 paf 9: #include "pa_exception.h"
1.154 paf 10: #include "pa_hash.h"
1.14 paf 11: #include "pa_globals.h"
1.154 paf 12: #include "pa_charsets.h"
1.214 paf 13: #include "pa_http.h"
1.223 misha 14: #include "pa_request_charsets.h"
1.241 misha 15: #include "pa_request.h"
1.98 paf 16:
1.283 moko 17: #include "pa_idna.h"
18: #include "pa_convert_utf.h"
19:
1.273 moko 20: #ifdef _MSC_VER
1.276 moko 21: #include <windows.h>
1.273 moko 22: #include <direct.h>
23: #endif
24:
1.341 ! moko 25: volatile const char * IDENT_PA_COMMON_C="$Id: pa_common.C,v 1.340 2025/07/04 00:18:35 moko Exp $" IDENT_PA_COMMON_H IDENT_PA_HASH_H IDENT_PA_ARRAY_H IDENT_PA_STACK_H;
1.267 moko 26:
1.93 paf 27: // some maybe-undefined constants
28:
1.82 paf 29: #ifndef _O_TEXT
30: # define _O_TEXT 0
31: #endif
32: #ifndef _O_BINARY
33: # define _O_BINARY 0
1.47 paf 34: #endif
1.80 paf 35:
1.138 paf 36: #ifdef HAVE_FTRUNCATE
37: # define PA_O_TRUNC 0
38: #else
39: # ifdef _O_TRUNC
40: # define PA_O_TRUNC _O_TRUNC
41: # else
42: # error you must have either ftruncate function or _O_TRUNC bit declared
43: # endif
1.154 paf 44: #endif
1.176 paf 45:
1.154 paf 46: // defines for globals
47:
48: #define FILE_STATUS_NAME "status"
49:
50: // globals
51:
52: const String file_status_name(FILE_STATUS_NAME);
53:
1.305 moko 54: String sql_bind_name(SQL_BIND_NAME);
55: String sql_limit_name(PA_SQL_LIMIT_NAME);
56: String sql_offset_name(PA_SQL_OFFSET_NAME);
57: String sql_default_name(SQL_DEFAULT_NAME);
58: String sql_distinct_name(SQL_DISTINCT_NAME);
59: String sql_value_type_name(SQL_VALUE_TYPE_NAME);
60:
1.301 moko 61: // forwards
62:
63: const UTF16* pa_utf16_encode(const char* in, Charset& source_charset);
64:
1.154 paf 65: // functions
1.127 paf 66:
1.301 moko 67: #ifdef _MSC_VER
68:
1.339 moko 69: #define PA_UTF16_ENC(value) (const wchar_t *)pa_utf16_encode(value, pa_thread_request().charsets.source())
70:
1.301 moko 71: int pa_stat(const char *pathname, struct stat *buffer){
1.339 moko 72: return _wstat64(PA_UTF16_ENC(pathname), buffer);
1.301 moko 73: }
74:
75: int pa_open(const char *pathname, int flags, int mode){
1.339 moko 76: return _wopen(PA_UTF16_ENC(pathname), flags, mode);
1.301 moko 77: }
78:
79: FILE *pa_fopen(const char *pathname, const char *mode){
1.339 moko 80: return _wfopen(PA_UTF16_ENC(pathname), PA_UTF16_ENC(mode));
1.301 moko 81: }
82:
1.338 moko 83: int pa_mkdir(const char *pathname, int){
1.339 moko 84: return _wmkdir(PA_UTF16_ENC(pathname));
1.338 moko 85: }
86:
87: int pa_rmdir(const char *pathname){
1.339 moko 88: return _wrmdir(PA_UTF16_ENC(pathname));
1.338 moko 89: }
90:
91: int pa_rename(const char *oldpath, const char *newpath){
1.340 moko 92: return _wrename(PA_UTF16_ENC(oldpath), PA_UTF16_ENC(newpath));
1.338 moko 93: }
94:
1.337 moko 95: int pa_unlink(const char *pathname){
1.339 moko 96: return _wunlink(PA_UTF16_ENC(pathname));
1.337 moko 97: }
98:
99: #else
100:
1.338 moko 101: #define pa_mkdir mkdir
102: #define pa_rmdir rmdir
103: #define pa_rename rename
1.337 moko 104: #define pa_unlink unlink
105:
1.301 moko 106: #endif
107:
1.206 paf 108: /// these options were handled but not checked elsewhere, now check them
1.239 misha 109: int pa_get_valid_file_options_count(HashStringValue& options) {
1.206 paf 110: int result=0;
111: if(options.get(PA_SQL_LIMIT_NAME))
112: result++;
113: if(options.get(PA_SQL_OFFSET_NAME))
114: result++;
115: if(options.get(PA_COLUMN_SEPARATOR_NAME))
116: result++;
117: if(options.get(PA_COLUMN_ENCLOSER_NAME))
118: result++;
1.223 misha 119: if(options.get(PA_CHARSET_NAME))
120: result++;
1.206 paf 121: return result;
122: }
123:
1.123 paf 124: #ifndef DOXYGEN
125: struct File_read_action_info {
1.154 paf 126: char **data; size_t *data_size;
1.318 moko 127: char* buf; uint64_t offset; size_t limit;
1.126 paf 128: };
1.123 paf 129: #endif
1.271 moko 130:
1.289 moko 131: static void file_read_action(struct stat& finfo, int f, const String& file_spec, void *context) {
1.310 moko 132: File_read_action_info& info = *static_cast<File_read_action_info *>(context);
1.326 moko 133: size_t to_read_size = check_file_size(info.limit && info.limit < (size_t)finfo.st_size ? info.limit : (size_t)finfo.st_size, &file_spec);
1.271 moko 134: if(to_read_size) {
1.320 moko 135: if(info.offset)
136: pa_lseek(f, info.offset, SEEK_SET); // seek never fails as POSIX allows the file offset to be set beyond the EOF
1.310 moko 137: *info.data = info.buf ? info.buf : (char *)pa_malloc_atomic(to_read_size+1);
138: ssize_t result = read(f, *info.data, to_read_size);
1.271 moko 139: if(result<0)
140: throw Exception("file.read", &file_spec, "read failed: %s (%d)", strerror(errno), errno);
1.310 moko 141: *info.data_size = result;
1.123 paf 142: } else { // empty file
1.209 paf 143: // for both, text and binary: for text we need that terminator, for binary we need nonzero pointer to be able to save such files
1.310 moko 144: *info.data = (char *)pa_malloc_atomic(1);
145: *(char*)(*info.data) = 0;
146: *info.data_size = 0;
1.123 paf 147: return;
148: }
1.126 paf 149: }
1.241 misha 150:
1.318 moko 151: File_read_result file_read_binary(const String& file_spec, bool fail_on_read_problem, char* buf, uint64_t offset, size_t limit) {
1.310 moko 152: File_read_result result = {false, 0, 0, 0};
153: File_read_action_info info = {&result.str, &result.length, buf, offset, limit};
1.309 moko 154:
1.310 moko 155: result.success = file_read_action_under_lock(file_spec, "read", file_read_action, &info, 0, fail_on_read_problem);
1.309 moko 156: return result;
157: }
158:
1.154 paf 159: File_read_result file_read(Request_charsets& charsets, const String& file_spec,
1.310 moko 160: bool as_text, HashStringValue *options,
1.229 misha 161: bool fail_on_read_problem,
1.310 moko 162: size_t offset = 0, size_t limit = 0, bool transcode_text_result = true) {
163: File_read_result result = {false, 0, 0, 0};
164: if(options){
165: int valid_options = pa_get_valid_file_options_count(*options);
166: if(valid_options != options->count())
1.262 misha 167: throw Exception(PARSER_RUNTIME, 0, CALLED_WITH_INVALID_OPTION);
1.241 misha 168: }
1.203 paf 169:
1.310 moko 170: File_read_action_info info = {&result.str, &result.length, 0, offset, limit};
1.161 paf 171:
1.310 moko 172: result.success = file_read_action_under_lock(file_spec, "read", file_read_action, &info, as_text, fail_on_read_problem);
1.223 misha 173:
1.241 misha 174: if(as_text){
175: if(result.success){
1.310 moko 176: Charset* asked_charset = 0;
177: if(options)
178: if(Value* vcharset_name = options->get(PA_CHARSET_NAME))
179: asked_charset = &pa_charsets.get(vcharset_name->as_string());
1.263 misha 180:
1.310 moko 181: asked_charset = pa_charsets.checkBOM(result.str, result.length, asked_charset);
1.287 moko 182:
1.263 misha 183: if(result.length && transcode_text_result && asked_charset){ // length must be checked because transcode returns CONST string in case length==0, which contradicts hacking few lines below
1.310 moko 184: String::C body = String::C(result.str, result.length);
1.263 misha 185: body=Charset::transcode(body, *asked_charset, charsets.source());
1.236 misha 186:
1.310 moko 187: result.str = const_cast<char*>(body.str); // hacking a little
188: result.length = body.length;
1.131 paf 189: }
190: }
1.241 misha 191: if(result.length)
192: fix_line_breaks(result.str, result.length);
1.123 paf 193: }
1.241 misha 194:
195: return result;
196: }
197:
198: File_read_result file_load(Request& r, const String& file_spec,
1.310 moko 199: bool as_text, HashStringValue *options,
1.241 misha 200: bool fail_on_read_problem,
1.310 moko 201: bool transcode_text_result) {
202:
203: size_t offset = 0;
204: size_t limit = 0;
205:
206: if(options){
207: if(Value *voffset = (Value *)options->get(sql_offset_name))
208: offset = r.process(*voffset).as_int();
209: if(Value *vlimit = (Value *)options->get(sql_limit_name))
210: limit = r.process(*vlimit).as_int();
211: // no check on options count here
212: }
1.241 misha 213:
214: if(file_spec.starts_with("http://")) {
1.310 moko 215: if(offset || limit)
1.289 moko 216: throw Exception(PARSER_RUNTIME, 0, "offset and load options are not supported for HTTP:// file load");
1.241 misha 217:
218: // fail on read problem
1.310 moko 219: File_read_http_result http = pa_internal_file_read_http(r, file_spec, as_text, options, transcode_text_result);
220:
221: File_read_result result = {true, http.str, http.length, http.headers};
222: return result;
1.241 misha 223: } else
1.310 moko 224: return file_read(r.charsets, file_spec, as_text, options, fail_on_read_problem, offset, limit, transcode_text_result);
225: }
1.126 paf 226:
1.310 moko 227: char* file_read_text(Request_charsets& charsets, const String& file_spec, bool fail_on_read_problem) {
228: File_read_result file = file_read(charsets, file_spec, true, 0, fail_on_read_problem);
229: return file.success ? file.str : 0;
1.123 paf 230: }
231:
1.310 moko 232: char* file_load_text(Request& r, const String& file_spec, bool fail_on_read_problem, HashStringValue* options, bool transcode_result) {
233: File_read_result file = file_load(r, file_spec, true, options, fail_on_read_problem, transcode_result);
234: return file.success ? file.str : 0;
235: }
1.257 pretende 236:
1.154 paf 237: #ifdef PA_SAFE_MODE
1.259 misha 238: void check_safe_mode(struct stat finfo, const String& file_spec, const char* fname) {
1.154 paf 239: if(finfo.st_uid/*foreign?*/!=geteuid()
240: && finfo.st_gid/*foreign?*/!=getegid())
1.289 moko 241: throw Exception(PARSER_RUNTIME,
242: &file_spec,
243: "parser is in safe mode: reading files of foreign group and user disabled "
244: "[recompile parser with --disable-safe-mode configure option], "
245: "actual filename '%s', fuid(%d)!=euid(%d) or fgid(%d)!=egid(%d)",
246: fname, finfo.st_uid, geteuid(), finfo.st_gid, getegid()
247: );
1.259 misha 248: }
249: #else
250: void check_safe_mode(struct stat, const String&, const char*) {
251: }
1.257 pretende 252: #endif
1.259 misha 253:
1.257 pretende 254:
1.335 moko 255: bool file_read_action_under_lock(const String& file_spec, const char* action_name, File_read_action action, void *context,
256: bool as_text, bool fail_on_read_problem) {
1.149 paf 257:
1.335 moko 258: const char* fname=file_spec.taint_cstr(String::L_FILE_SPEC);
1.33 paf 259: int f;
260:
261: // first open, next stat:
1.45 paf 262: // directory update of NTFS hard links performed on open.
1.33 paf 263: // ex:
264: // a.html:^test[] and b.html hardlink to a.html
265: // user inserts ! before ^test in a.html
1.126 paf 266: // directory entry of b.html in NTFS not updated at once,
1.35 paf 267: // they delay update till open, so we would receive "!^test[" string
268: // if would do stat, next open.
1.123 paf 269: // later: it seems, even this does not help sometimes
1.335 moko 270: if((f=pa_open(fname, O_RDONLY | (as_text ? _O_TEXT : _O_BINARY) ))>=0) {
1.123 paf 271: try {
1.335 moko 272: int pa_errno=pa_lock_shared_blocking(f);
273: if(pa_errno!=0)
274: throw Exception("file.lock", &file_spec, "shared lock failed: %s (%d), actual filename '%s'", strerror(pa_errno), pa_errno, fname);
1.123 paf 275:
1.124 paf 276: struct stat finfo;
1.298 moko 277: if(pa_fstat(f, &finfo)!=0)
1.124 paf 278: throw Exception("file.missing", // hardly possible: we just opened it OK
1.289 moko 279: &file_spec, "stat failed: %s (%d), actual filename '%s'", strerror(errno), errno, fname);
1.124 paf 280:
1.149 paf 281: check_safe_mode(finfo, file_spec, fname);
1.32 paf 282:
1.289 moko 283: action(finfo, f, file_spec, context);
1.123 paf 284: } catch(...) {
1.162 paf 285: pa_unlock(f);close(f);
1.123 paf 286: if(fail_on_read_problem)
1.154 paf 287: rethrow;
1.289 moko 288: return false;
1.123 paf 289: }
1.87 paf 290:
1.162 paf 291: pa_unlock(f);close(f);
1.72 parser 292: return true;
1.229 misha 293: } else {
1.118 paf 294: if(fail_on_read_problem)
1.289 moko 295: throw Exception(errno==EACCES ? "file.access" : (errno==ENOENT || errno==ENOTDIR || errno==ENODEV) ? "file.missing" : 0,
296: &file_spec, "%s failed: %s (%d), actual filename '%s'", action_name, strerror(errno), errno, fname);
1.118 paf 297: return false;
298: }
1.8 paf 299: }
300:
1.202 paf 301: void create_dir_for_file(const String& file_spec) {
1.324 moko 302: const char *str=file_spec.taint_cstr(String::L_FILE_SPEC);
303: if(str[0]){
304: const char *pos=str+1;
1.339 moko 305: while((pos=strpbrk(pos, "/\\")) && pos[1]) { // to avoid trailing /, see #1166
1.324 moko 306: pa_mkdir(pa_strdup(str,pos-str), 0775);
307: pos++;
308: }
1.63 parser 309: }
310: }
311:
1.335 moko 312: bool file_write_action_under_lock(const String& file_spec, const char* action_name, File_write_action action, void *context,
313: bool as_text, bool do_append, bool do_block, bool fail_on_lock_problem) {
314:
1.247 misha 315: const char* fname=file_spec.taint_cstr(String::L_FILE_SPEC);
1.28 paf 316: int f;
1.80 paf 317: if(access(fname, W_OK)!=0) // no
1.126 paf 318: create_dir_for_file(file_spec);
1.50 paf 319:
1.335 moko 320: if((f=pa_open(fname, O_CREAT | O_RDWR | (as_text ? _O_TEXT : _O_BINARY) | (do_append ? O_APPEND : PA_O_TRUNC), 0664))>=0) {
321: int pa_errno=do_block ? pa_lock_exclusive_blocking(f) : pa_lock_exclusive_nonblocking(f);
322: if(pa_errno!=0) {
323: Exception e("file.lock", &file_spec, "shared lock failed: %s (%d), actual filename '%s'", strerror(pa_errno), pa_errno, fname);
324: close(f);
1.110 paf 325: if(fail_on_lock_problem)
326: throw e;
1.98 paf 327: return false;
328: }
1.96 paf 329:
1.158 paf 330: try {
1.254 misha 331: #if (defined(HAVE_FCHMOD) && defined(PA_SAFE_MODE))
332: struct stat finfo;
1.298 moko 333: if(pa_fstat(f, &finfo)==0 && finfo.st_mode & 0111)
1.254 misha 334: fchmod(f, finfo.st_mode & 0666/*clear executable bits*/); // backward: ignore errors if any
335: #endif
336: action(f, context);
1.158 paf 337: } catch(...) {
1.138 paf 338: #ifdef HAVE_FTRUNCATE
1.104 paf 339: if(!do_append)
1.336 moko 340: PA_UNUSED int ignore_result=ftruncate(f, lseek(f, 0, SEEK_CUR)); // one cannot use O_TRUNC, read lower
1.138 paf 341: #endif
1.336 moko 342: pa_unlock(f);close(f);
1.154 paf 343: rethrow;
1.158 paf 344: }
1.80 paf 345:
1.138 paf 346: #ifdef HAVE_FTRUNCATE
1.104 paf 347: if(!do_append)
1.336 moko 348: PA_UNUSED int ignore_result=ftruncate(f, lseek(f, 0, SEEK_CUR)); // O_TRUNC truncates even exclusevely write-locked file [thanks to Igor Milyakov <virtan@rotabanner.com> for discovering]
1.138 paf 349: #endif
1.336 moko 350: pa_unlock(f);close(f);
1.98 paf 351: return true;
1.80 paf 352: } else
1.289 moko 353: throw Exception(errno==EACCES ? "file.access" : 0, &file_spec, "%s failed: %s (%d), actual filename '%s'", action_name, strerror(errno), errno, fname);
1.96 paf 354: // here should be nothing, see rethrow above
355: }
356:
357: #ifndef DOXYGEN
358: struct File_write_action_info {
1.250 misha 359: const char* str;
360: size_t length;
1.126 paf 361: };
1.96 paf 362: #endif
1.271 moko 363:
1.96 paf 364: static void file_write_action(int f, void *context) {
1.126 paf 365: File_write_action_info& info=*static_cast<File_write_action_info *>(context);
1.154 paf 366: if(info.length) {
1.271 moko 367: ssize_t written=write(f, info.str, info.length);
1.116 paf 368: if(written<0)
1.271 moko 369: throw Exception("file.write", 0, "write failed: %s (%d)", strerror(errno), errno);
1.275 moko 370: if((size_t)written!=info.length)
1.271 moko 371: throw Exception("file.write", 0, "write failed: %u of %u bytes written", written, info.length);
1.113 paf 372: }
1.96 paf 373: }
1.271 moko 374:
1.96 paf 375: void file_write(
1.250 misha 376: Request_charsets& charsets,
377: const String& file_spec,
378: const char* data,
379: size_t size,
1.126 paf 380: bool as_text,
1.250 misha 381: bool do_append,
382: Charset* asked_charset) {
383:
384: if(as_text && asked_charset){
385: String::C body=String::C(data, size);
386: body=Charset::transcode(body, charsets.source(), *asked_charset);
387: data=body.str;
388: size=body.length;
389: };
390:
1.126 paf 391: File_write_action_info info={data, size};
1.225 misha 392:
1.98 paf 393: file_write_action_under_lock(
1.154 paf 394: file_spec,
1.225 misha 395: "write",
396: file_write_action,
397: &info,
1.154 paf 398: as_text,
399: do_append);
1.30 paf 400: }
401:
1.261 misha 402: static size_t get_dir(char* fname, size_t helper_length){
403: bool dir=false;
404: size_t pos=0;
405: for(pos=helper_length; pos; pos--){
406: char c=fname[pos-1];
407: if(c=='/' || c=='\\'){
408: fname[pos-1]=0;
409: dir=true;
410: } else if(dir) break;
411: }
412: return pos;
413: }
414:
1.312 moko 415: bool entry_exists(const char* fname, struct stat *afinfo) {
416: struct stat lfinfo;
417: bool result=pa_stat(fname, &lfinfo)==0;
418: if(afinfo)
419: *afinfo=lfinfo;
420: return result;
421: }
422:
423: bool entry_exists(const String& file_spec) {
424: return entry_exists(file_spec.taint_cstr(String::L_FILE_SPEC), 0);
425: }
426:
427: static bool entry_ifdir(char *fname, bool need_dir) {
1.261 misha 428: if(need_dir){
429: size_t size=strlen(fname);
430: while(size) {
431: char c=fname[size-1];
432: if(c=='/' || c=='\\')
433: fname[--size]=0;
434: else
435: break;
436: }
437: }
438:
439: struct stat finfo;
1.311 moko 440: if(entry_exists(fname, &finfo)) {
1.261 misha 441: bool is_dir=(finfo.st_mode&S_IFDIR) != 0;
442: return is_dir==need_dir;
443: }
444: return false;
445: }
446:
1.312 moko 447: static bool entry_ifdir(const String& file_spec, bool need_dir) {
448: return entry_ifdir(file_spec.taint_cstrm(String::L_FILE_SPEC), need_dir);
449: }
450:
1.63 parser 451: // throws nothing! [this is required in file_move & file_delete]
1.277 moko 452: static void rmdir(const String& file_spec, size_t pos_after) {
1.261 misha 453: char* dir_spec=file_spec.taint_cstrm(String::L_FILE_SPEC);
454: size_t length=strlen(dir_spec);
455: while( (length=get_dir(dir_spec, length)) && (length > pos_after) ){
1.274 moko 456: #ifdef _MSC_VER
1.312 moko 457: if(!entry_ifdir(dir_spec, true))
1.261 misha 458: break;
1.339 moko 459: DWORD attrs=GetFileAttributesW(PA_UTF16_ENC(dir_spec));
1.261 misha 460: if(
461: (attrs==INVALID_FILE_ATTRIBUTES)
462: || !(attrs & FILE_ATTRIBUTE_DIRECTORY)
463: || (attrs & FILE_ATTRIBUTE_REPARSE_POINT)
464: )
465: break;
466: #endif
1.338 moko 467: if(pa_rmdir(dir_spec))
1.261 misha 468: break;
469: };
1.50 paf 470: }
1.239 misha 471:
1.269 misha 472: bool file_delete(const String& file_spec, bool fail_on_problem, bool keep_empty_dirs) {
1.247 misha 473: const char* fname=file_spec.taint_cstr(String::L_FILE_SPEC);
1.337 moko 474: if(pa_unlink(fname)!=0) {
1.164 paf 475: if(fail_on_problem)
1.289 moko 476: throw Exception(errno==EACCES?"file.access":errno==ENOENT?"file.missing":0,
477: &file_spec, "unlink failed: %s (%d), actual filename '%s'", strerror(errno), errno, fname);
1.93 paf 478: else
479: return false;
1.282 moko 480: }
1.50 paf 481:
1.269 misha 482: if(!keep_empty_dirs)
483: rmdir(file_spec, 1);
484:
1.93 paf 485: return true;
1.60 parser 486: }
1.239 misha 487:
1.269 misha 488: void file_move(const String& old_spec, const String& new_spec, bool keep_empty_dirs) {
1.247 misha 489: const char* old_spec_cstr=old_spec.taint_cstr(String::L_FILE_SPEC);
490: const char* new_spec_cstr=new_spec.taint_cstr(String::L_FILE_SPEC);
1.63 parser 491:
1.126 paf 492: create_dir_for_file(new_spec);
1.63 parser 493:
1.338 moko 494: if(pa_rename(old_spec_cstr, new_spec_cstr)!=0)
1.289 moko 495: throw Exception(errno==EACCES ? "file.access" : errno==ENOENT ? "file.missing" : 0,
496: &old_spec, "rename failed: %s (%d), actual filename '%s' to '%s'", strerror(errno), errno, old_spec_cstr, new_spec_cstr);
1.63 parser 497:
1.269 misha 498: if(!keep_empty_dirs)
499: rmdir(old_spec, 1);
1.31 paf 500: }
501:
1.51 paf 502:
1.215 paf 503: bool file_exist(const String& file_spec) {
1.312 moko 504: return entry_ifdir(file_spec, false);
1.51 paf 505: }
1.239 misha 506:
1.215 paf 507: bool dir_exists(const String& file_spec) {
1.312 moko 508: return entry_ifdir(file_spec, true);
1.65 parser 509: }
1.239 misha 510:
1.215 paf 511: const String* file_exist(const String& path, const String& name) {
1.154 paf 512: String& result=*new String(path);
1.270 moko 513: if(path.last_char() != '/')
514: result << "/";
1.154 paf 515: result << name;
1.215 paf 516: return file_exist(result)?&result:0;
1.43 paf 517: }
1.239 misha 518:
1.43 paf 519: bool file_executable(const String& file_spec) {
1.247 misha 520: return access(file_spec.taint_cstr(String::L_FILE_SPEC), X_OK)==0;
1.44 paf 521: }
522:
1.296 moko 523: bool file_stat(const String& file_spec, uint64_t& rsize, time_t& ratime, time_t& rmtime, time_t& rctime, bool fail_on_read_problem) {
1.247 misha 524: const char* fname=file_spec.taint_cstr(String::L_FILE_SPEC);
1.154 paf 525: struct stat finfo;
1.298 moko 526: if(pa_stat(fname, &finfo)!=0) {
1.64 parser 527: if(fail_on_read_problem)
1.289 moko 528: throw Exception("file.missing", &file_spec, "getting file size failed: %s (%d), real filename '%s'", strerror(errno), errno, fname);
1.64 parser 529: else
530: return false;
1.282 moko 531: }
1.58 parser 532: rsize=finfo.st_size;
1.299 moko 533: ratime=(time_t)finfo.st_atime;
534: rmtime=(time_t)finfo.st_mtime;
535: rctime=(time_t)finfo.st_ctime;
1.64 parser 536: return true;
1.18 paf 537: }
538:
1.313 moko 539: size_t check_file_size(uint64_t size, const String* file_spec){
1.319 moko 540: if(size > (uint64_t)pa_file_size_limit)
1.313 moko 541: throw Exception(PARSER_RUNTIME, file_spec, "content size of %.15g bytes exceeds the limit (%.15g bytes)", (double)size, (double)pa_file_size_limit);
1.300 moko 542: return (size_t)size;
543: }
544:
1.278 moko 545: /**
546: String related functions
547: */
548:
549: bool capitalized(const char* s){
550: bool upper=true;
551: for(const char* c=s; *c; c++){
552: if(*c != (upper ? toupper((unsigned char)*c) : tolower((unsigned char)*c)))
553: return false;
554: upper=strchr("-_ ", *c) != 0;
555: }
556: return true;
557: }
558:
559: const char* capitalize(const char* s){
560: if(!s || capitalized(s))
561: return s;
562:
563: char* result=pa_strdup(s);
564: if(result){
565: bool upper=true;
566: for(char* c=result; *c; c++){
567: *c=upper ? (char)toupper((unsigned char)*c) : (char)tolower((unsigned char)*c);
568: upper=strchr("-_ ", *c) != 0;
569: }
570: }
571: return (const char*)result;
572: }
573:
1.323 moko 574: char *str_lower(const char *s, size_t length){
575: char *result=pa_strdup(s, length);
1.290 moko 576: for(char* c=result; *c; c++)
577: *c=(char)tolower((unsigned char)*c);
578: return result;
579: }
580:
1.323 moko 581: char *str_upper(const char *s, size_t length){
582: char *result=pa_strdup(s, length);
1.290 moko 583: for(char* c=result; *c; c++)
584: *c=(char)toupper((unsigned char)*c);
585: return result;
586: }
587:
1.278 moko 588: void fix_line_breaks(char *str, size_t& length) {
589: //_asm int 3;
590: const char* const eob=str+length;
591: char* dest=str;
592: // fix DOS: \r\n -> \n
593: // fix Macintosh: \r -> \n
594: char* bol=str;
595: while(char* eol=(char*)memchr(bol, '\r', eob -bol)) {
596: size_t len=eol-bol;
597: if(dest!=bol)
598: memmove(dest, bol, len);
599: dest+=len;
600: *dest++='\n';
601:
602: if(&eol[1]<eob && eol[1]=='\n') { // \r, \n = DOS
603: bol=eol+2;
604: length--;
605: } else // \r, not \n = Macintosh
606: bol=eol+1;
607: }
608: // last piece without \r
609: if(dest!=bol)
610: memmove(dest, bol, eob-bol);
611: str[length]=0; // terminating
612: }
613:
614: /**
615: scans for @a delim[default \n] in @a *row_ref,
616: @return piece of line before it or end of string, if no @a delim found
617: assigns @a *row_ref to point right after delimiter if there were one
618: or to zero if no @a delim were found.
619: */
620:
1.126 paf 621: char* getrow(char* *row_ref, char delim) {
1.229 misha 622: char* result=*row_ref;
623: if(result) {
1.126 paf 624: *row_ref=strchr(result, delim);
1.8 paf 625: if(*row_ref)
626: *((*row_ref)++)=0;
627: else if(!*result)
628: return 0;
1.229 misha 629: }
630: return result;
1.8 paf 631: }
632:
1.126 paf 633: char* lsplit(char* string, char delim) {
1.229 misha 634: if(string) {
1.126 paf 635: char* v=strchr(string, delim);
1.8 paf 636: if(v) {
637: *v=0;
638: return v+1;
639: }
1.229 misha 640: }
641: return 0;
1.8 paf 642: }
643:
1.126 paf 644: char* lsplit(char* *string_ref, char delim) {
1.229 misha 645: char* result=*string_ref;
1.126 paf 646: char* next=lsplit(*string_ref, delim);
1.229 misha 647: *string_ref=next;
648: return result;
1.9 paf 649: }
650:
1.126 paf 651: char* rsplit(char* string, char delim) {
1.229 misha 652: if(string) {
1.126 paf 653: char* v=strrchr(string, delim);
1.18 paf 654: if(v) {
1.9 paf 655: *v=0;
656: return v+1;
657: }
1.229 misha 658: }
1.314 moko 659: return NULL;
660: }
661:
1.325 moko 662: void pa_strncpy(char *dst, const char *src, size_t n){
663: size_t left = n;
664:
665: if (left != 0 && src) {
666: while (--left != 0) {
667: if ((*dst++ = *src++) == '\0')
668: return;
669: }
670: }
671: if (n != 0)
672: *dst = '\0';
673: }
674:
1.314 moko 675: #define STRCAT_STEP(str, len) if(str) { memcpy(ptr, str, len); ptr += len; }
676:
677: char *pa_strcat(const char *a, const char *b, const char *c) {
678: size_t len_a = a ? strlen(a) : 0;
1.315 moko 679: size_t len_b = b ? strlen(b) : 0;
680: size_t len_c = c ? strlen(c) : 0;
1.314 moko 681: char *result=new(PointerFreeGC) char[len_a + len_b + len_c +1/*0*/];
682: char *ptr=result;
683: STRCAT_STEP(a, len_a);
684: STRCAT_STEP(b, len_b);
685: STRCAT_STEP(c, len_c);
686: *ptr='\0';
687: return result;
1.10 paf 688: }
689:
1.317 moko 690: const char *pa_filename(const char *path){
691: if(!path)
692: return NULL;
693: for(const char *c = path + strlen(path) - 1; c >= path; c--) {
694: if(*c == '/' || *c == '\\')
695: return c+1;
696: }
697: return path;
698: }
1.229 misha 699:
1.36 paf 700: size_t stdout_write(const void *buf, size_t size) {
1.12 paf 701: #ifdef WIN32
1.187 paf 702: size_t to_write = size;
1.12 paf 703: do{
1.316 moko 704: int chunk_written=fwrite(buf, 1, min((size_t)8*0x400, size), stdout);
1.12 paf 705: if(chunk_written<=0)
706: break;
707: size-=chunk_written;
1.36 paf 708: buf=((const char*)buf)+chunk_written;
1.316 moko 709: } while(size>0);
1.12 paf 710:
1.316 moko 711: fflush(stdout);
1.187 paf 712: return to_write-size;
1.12 paf 713: #else
1.316 moko 714: size_t result=fwrite(buf, 1, size, stdout);
715: fflush(stdout);
716: return result;
1.12 paf 717: #endif
1.2 paf 718: }
1.14 paf 719:
1.229 misha 720: enum EscapeState {
1.341 ! moko 721: EscapeRest,
! 722: EscapeFirst,
1.229 misha 723: EscapeSecond,
724: EscapeUnicode
725: };
726:
1.236 misha 727: // @todo prescan for reduce required size (unescaped sting in 1 byte charset requires less memory usually)
1.258 misha 728: char* unescape_chars(const char* cp, int len, Charset* charset, bool js){
1.236 misha 729: char* s=new(PointerFreeGC) char[len+1]; // must be enough (%uXXXX==6 bytes, max utf-8 char length==6 bytes)
1.336 moko 730: XMLByte* dst=(XMLByte *)s;
1.229 misha 731: EscapeState escapeState=EscapeRest;
732: uint escapedValue=0;
733: int srcPos=0;
1.230 misha 734: short int jsCnt=0;
1.236 misha 735: while(srcPos<len){
1.229 misha 736: uchar c=(uchar)cp[srcPos];
1.258 misha 737: if(c=='%' || (c=='\\' && js)){
1.229 misha 738: escapeState=EscapeFirst;
739: } else {
740: switch(escapeState) {
741: case EscapeRest:
1.286 moko 742: if(c=='+' && !js){
1.230 misha 743: *dst++=' ';
1.229 misha 744: } else {
1.230 misha 745: *dst++=c;
1.229 misha 746: }
747: break;
748: case EscapeFirst:
1.232 misha 749: if(charset && c=='u'){
1.229 misha 750: // escaped unicode value: %u0430
751: jsCnt=0;
752: escapedValue=0;
753: escapeState=EscapeUnicode;
754: } else {
1.231 misha 755: if(isxdigit(c)){
1.229 misha 756: escapedValue=hex_value[c] << 4;
757: escapeState=EscapeSecond;
758: } else {
1.230 misha 759: *dst++=c;
1.229 misha 760: escapeState=EscapeRest;
761: }
762: }
763: break;
764: case EscapeSecond:
1.231 misha 765: if(isxdigit(c)){
1.229 misha 766: escapedValue+=hex_value[c];
1.230 misha 767: *dst++=(char)escapedValue;
1.229 misha 768: }
769: escapeState=EscapeRest;
770: break;
771: case EscapeUnicode:
1.231 misha 772: if(isxdigit(c)){
1.229 misha 773: escapedValue=(escapedValue << 4) + hex_value[c];
774: if(++jsCnt==4){
1.230 misha 775: // transcode utf8 char to client charset (we can lost some chars here)
1.336 moko 776: charset->store_Char(dst, (XMLCh)escapedValue, '?');
1.229 misha 777: escapeState=EscapeRest;
778: }
779: } else {
780: // not full unicode value
781: escapeState=EscapeRest;
782: }
783: break;
784: }
785: }
786:
787: srcPos++;
788: }
789:
1.230 misha 790: *dst=0; // zero-termination
1.229 misha 791: return s;
792: }
1.24 paf 793:
1.268 misha 794: char *search_stop(char*& current, char cstop_at) {
795: // sanity check
796: if(!current)
797: return 0;
798:
799: // skip leading WS
800: while(*current==' ' || *current=='\t')
801: current++;
802: if(!*current)
803: return current=0;
804:
805: char *result=current;
806: if(char *pstop_at=strchr(current, cstop_at)) {
807: *pstop_at=0;
808: current=pstop_at+1;
809: } else
810: current=0;
811: return result;
812: }
813:
1.24 paf 814: #ifdef WIN32
1.126 paf 815: void back_slashes_to_slashes(char* s) {
1.24 paf 816: if(s)
817: for(; *s; s++)
818: if(*s=='\\')
1.126 paf 819: *s='/';
1.24 paf 820: }
821: #endif
1.41 paf 822:
1.232 misha 823: size_t strpos(const char *str, const char *substr) {
824: const char *p = strstr(str, substr);
825: return (p==0)?STRING_NOT_FOUND:p-str;
826: }
827:
1.326 moko 828: size_t remove_crlf(char* start, char* end) {
1.226 misha 829: char* from=start;
830: char* to=start;
831: bool skip=false;
832: while(from < end){
833: switch(*from){
834: case '\n':
835: case '\r':
836: case '\t':
837: case ' ':
838: if(!skip){
839: *to=' ';
840: to++;
841: skip=true;
842: }
843: break;
844: default:
845: if(from != to)
846: *to=*from;
847: to++;
848: skip=false;
1.69 parser 849: }
1.226 misha 850: from++;
851: }
852: return to-start;
1.91 paf 853: }
854:
1.279 moko 855: const char* hex_digits="0123456789ABCDEF";
856:
1.278 moko 857: const char* hex_string(unsigned char* bytes, size_t size, bool upcase) {
858: char *bytes_hex=new(PointerFreeGC) char [size*2/*byte->hh*/+1/*for zero-teminator*/];
859: unsigned char *src=bytes;
860: unsigned char *end=bytes+size;
861: char *dest=bytes_hex;
862:
1.279 moko 863: const char *hex=upcase? hex_digits : "0123456789abcdef";
1.278 moko 864:
865: for(; src<end; src++) {
866: *dest++=hex[*src/0x10];
867: *dest++=hex[*src%0x10];
868: }
869: *dest=0;
870:
871: return bytes_hex;
872: }
1.91 paf 873:
1.321 moko 874: ssize_t file_block_read(const int f, void* buffer, const size_t size){
875: ssize_t nCount = read(f, buffer, size);
1.221 misha 876: if (nCount < 0)
1.289 moko 877: throw Exception("file.read", 0, "read failed: %s (%d)", strerror(errno), errno);
1.221 misha 878: return nCount;
879: }
880:
1.278 moko 881: static unsigned long crc32Table[256];
882: static void InitCrc32Table()
883: {
884: if(crc32Table[1] == 0){
885: // This is the official polynomial used by CRC32 in PKZip.
886: // Often times the polynomial shown reversed as 0x04C11DB7.
887: static const unsigned long dwPolynomial = 0xEDB88320;
888:
889: for(int i = 0; i < 256; i++)
890: {
891: unsigned long dwCrc = i;
892: for(int j = 8; j > 0; j--)
893: {
894: if(dwCrc & 1)
895: dwCrc = (dwCrc >> 1) ^ dwPolynomial;
896: else
897: dwCrc >>= 1;
898: }
899: crc32Table[i] = dwCrc;
900: }
901: }
902: }
903:
904: inline void CalcCrc32(const unsigned char byte, unsigned long &crc32)
905: {
906: crc32 = ((crc32) >> 8) ^ crc32Table[(byte) ^ ((crc32) & 0x000000FF)];
907: }
908:
909:
1.297 moko 910: unsigned long pa_crc32(const char *in, size_t in_size){
1.218 misha 911: unsigned long crc32=0xFFFFFFFF;
1.220 misha 912:
1.240 misha 913: InitCrc32Table();
1.239 misha 914: for(size_t i = 0; i<in_size; i++)
915: CalcCrc32(in[i], crc32);
1.220 misha 916:
1.218 misha 917: return ~crc32;
918: }
919:
1.289 moko 920: static void file_crc32_file_action(struct stat& finfo, int f, const String&, void *context) {
1.218 misha 921: unsigned long& crc32=*static_cast<unsigned long *>(context);
922: if(finfo.st_size) {
923: InitCrc32Table();
1.220 misha 924: int nCount=0;
1.218 misha 925: do {
1.221 misha 926: unsigned char buffer[FILE_BUFFER_SIZE];
927: nCount = file_block_read(f, buffer, sizeof(buffer));
1.220 misha 928: for(int i = 0; i < nCount; i++) CalcCrc32(buffer[i], crc32);
929: } while(nCount > 0);
1.218 misha 930: }
931: }
932:
1.297 moko 933: unsigned long pa_crc32(const String& file_spec){
1.278 moko 934: unsigned long crc32=0xFFFFFFFF;
935: file_read_action_under_lock(file_spec, "crc32", file_crc32_file_action, &crc32);
936: return ~crc32;
937: }
938:
939: // content-type: xxx; charset=WE-NEED-THIS
940: // content-type: xxx; charset="WE-NEED-THIS"
941: // content-type: xxx; charset="WE-NEED-THIS";
942: Charset* detect_charset(const char* content_type){
943: if(content_type){
1.291 moko 944: char* CONTENT_TYPE=str_upper(content_type);
1.278 moko 945:
946: if(const char* begin=strstr(CONTENT_TYPE, "CHARSET=")){
947: begin+=8; // skip "CHARSET="
948: char* end=0;
949: if(*begin && (*begin=='"' || *begin =='\'')){
950: char quote=*begin;
951: begin++;
952: end=(char*)strchr(begin, quote);
953: }
954: if(!end)
955: end=(char*)strchr(begin, ';');
956:
957: if(end)
958: *end=0; // terminator
959:
1.295 moko 960: return *begin ? &pa_charsets.get_direct(begin) : 0;
1.278 moko 961: }
962: }
963: return 0;
964: }
965:
1.301 moko 966: const UTF16* pa_utf16_encode(const char* in, Charset& source_charset){
1.302 moko 967: if(!in)
968: return 0;
1.301 moko 969:
1.302 moko 970: String::C sIn(in,strlen(in));
1.301 moko 971:
1.302 moko 972: UTF16* utf16=(UTF16*)pa_malloc_atomic(sIn.length*2+2);
973: UTF16* utf16_end=utf16;
1.301 moko 974:
1.302 moko 975: if(!source_charset.isUTF8())
976: sIn=Charset::transcode(sIn, source_charset, pa_UTF8_charset);
1.301 moko 977:
1.302 moko 978: int status=pa_convertUTF8toUTF16((const UTF8**)&sIn.str, (const UTF8*)(sIn.str+sIn.length), &utf16_end, utf16+sIn.length, strictConversion);
979: if(status != conversionOK)
980: throw Exception("utf-16 encode", new String(in), "utf-16 conversion failed (%d)", status);
1.301 moko 981:
1.302 moko 982: *utf16_end=0;
1.301 moko 983:
1.302 moko 984: return utf16;
1.301 moko 985: }
986:
987: const char* pa_utf16_decode(const UTF16* in, Charset& asked_charset){
988: if(!in)
989: return 0;
990:
991: const UTF16* utf16_start=in;
992: const UTF16* utf16_end;
993:
994: for(utf16_end=in; *utf16_end; utf16_end++);
995:
996: char *result = (char *)pa_malloc_atomic((utf16_end-in)*6+1);
997: char *result_end = result;
998:
999: int status=pa_convertUTF16toUTF8(&utf16_start, utf16_end, (UTF8**)&result_end, (UTF8*)(result+(utf16_end-in)*6), strictConversion);
1000:
1001: if(status != conversionOK)
1002: throw Exception("utf-16 decode", 0, "utf conversion failed (%d)", status);
1003:
1004: *result_end='\0';
1005:
1.302 moko 1006: if(asked_charset.isUTF8())
1007: return result;
1.301 moko 1008:
1.302 moko 1009: return Charset::transcode(result, pa_UTF8_charset, asked_charset).cstr();
1.301 moko 1010: }
1.278 moko 1011:
1.283 moko 1012: static bool is_latin(const char *in){
1013: for(; *in; in++){
1014: if ((unsigned char)(*in) > 0x7F)
1015: return false;
1016: }
1017: return true;
1018: }
1019:
1020: #define MAX_IDNA_LENGTH 256
1021:
1.301 moko 1022: const char *pa_idna_encode(const char *in, Charset& source_charset){
1.283 moko 1023: if(!in || is_latin(in))
1024: return in;
1025:
1026: uint32_t utf32[MAX_IDNA_LENGTH];
1027: uint32_t *utf32_end=utf32;
1028:
1029: String::C sIn(in,strlen(in));
1030:
1031: if(!source_charset.isUTF8())
1.295 moko 1032: sIn=Charset::transcode(sIn, source_charset, pa_UTF8_charset);
1.283 moko 1033:
1034: int status=pa_convertUTF8toUTF32((const UTF8**)&sIn.str, (const UTF8*)(sIn.str+sIn.length), &utf32_end, utf32+MAX_IDNA_LENGTH-1, strictConversion);
1035: if(status != conversionOK)
1036: throw Exception("idna encode", new String(in), "utf conversion failed (%d)", status);
1037:
1038: *utf32_end=0;
1039:
1040: char *result = (char *)pa_malloc(MAX_IDNA_LENGTH);
1041: status=pa_idna_to_ascii_4z(utf32, result, MAX_IDNA_LENGTH, 0);
1042: if(status != IDNA_SUCCESS)
1043: throw Exception("idna encode", new String(in), "encode failed: %s", pa_idna_strerror(status));
1044:
1045: return result;
1046: }
1047:
1048: const char *pa_idna_decode(const char *in, Charset &asked_charset){
1049: if(!in || !(*in))
1050: return in;
1051:
1052: uint32_t utf32[MAX_IDNA_LENGTH];
1053: const uint32_t *utf32_start=utf32;
1054: uint32_t *utf32_end;
1055:
1056: int status=pa_idna_to_unicode_4z(in, utf32, MAX_IDNA_LENGTH, 0);
1057: if(status != IDNA_SUCCESS)
1058: throw Exception("idna decode", new String(in), "decode failed: %s", pa_idna_strerror(status));
1059:
1060: for(utf32_end=utf32; *utf32_end; utf32_end++);
1061:
1062: char *result = (char *)pa_malloc(MAX_IDNA_LENGTH);
1063: char *result_end = result;
1064:
1065: status=pa_convertUTF32toUTF8(&utf32_start, utf32_end, (UTF8**)&result_end, (UTF8*)(result+MAX_IDNA_LENGTH-1), strictConversion);
1066:
1067: if(status != conversionOK)
1068: throw Exception("idna decode", new String(in), "utf conversion failed (%d)", status);
1069:
1070: *result_end='\0';
1071:
1072: if(!asked_charset.isUTF8())
1.295 moko 1073: result = (char *)Charset::transcode(result, pa_UTF8_charset, asked_charset).cstr();
1.283 moko 1074:
1075: return result;
1076: }
1.292 moko 1077: /// must be last in this file
1078: #undef vsnprintf
1079: int pa_vsnprintf(char* b, size_t s, const char* f, va_list l) {
1080: if(!s)
1081: return 0;
1082:
1083: int r;
1084: // note: on win32 & maybe somewhere else
1085: // vsnprintf do not writes terminating 0 in 'buffer full' case, reducing
1086: // http://stackoverflow.com/questions/2915672/snprintf-and-visual-studio-2010
1087: --s;
1088:
1089: // clients do not check for negative 's', feature: ignore such prints
1090: if((ssize_t)s<0)
1091: return 0;
1092:
1093: #ifdef _MSC_VER
1094: // win32: if the number of bytes to write exceeds buffer, then count bytes are written and -1 is returned
1095: r=_vsnprintf(b, s, f, l);
1096: if(r<0)
1097: r=s;
1098: #else
1099: r=vsnprintf(b, s, f, l);
1100: /*
1101: solaris: man vsnprintf
1102:
1103: The snprintf() function returns the number of characters
1104: formatted, that is, the number of characters that would have
1105: been written to the buffer if it were large enough. If the
1106: value of n is 0 on a call to snprintf(), an unspecified
1107: value less than 1 is returned.
1108: */
1109:
1110: if(r<0)
1111: r=0;
1112: else if((size_t)r>s)
1113: r=s;
1114: #endif
1115: b[r]=0;
1116: return r;
1117: }
1118:
1119: int pa_snprintf(char* b, size_t s, const char* f, ...) {
1120: va_list l;
1121: va_start(l, f);
1122: int r=pa_vsnprintf(b, s, f, l);
1123: va_end(l);
1124: return r;
1125: }
1126:
E-mail:
![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to pa_common.C CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [parser3project] / parser3 / src / main