Annotation of parser3/src/main/pa_http.C, revision 1.72
1.1 paf 1: /** @file
2: Parser: http support functions.
3:
1.69 moko 4: Copyright (c) 2001-2015 Art. Lebedev Studio (http://www.artlebedev.com)
1.1 paf 5: Author: Alexandr Petrosian <paf@design.ru> (http://paf.design.ru)
6: */
7:
8: #include "pa_http.h"
9: #include "pa_common.h"
10: #include "pa_charsets.h"
11: #include "pa_request_charsets.h"
1.22 misha 12: #include "pa_request.h"
13: #include "pa_vfile.h"
14: #include "pa_random.h"
1.1 paf 15:
1.72 ! moko 16: volatile const char * IDENT_PA_HTTP_C="$Id: pa_http.C,v 1.71 2016/07/21 17:05:37 moko Exp $" IDENT_PA_HTTP_H;
1.59 moko 17:
18: #ifdef _MSC_VER
19: #include <windows.h>
20: #else
21: #define closesocket close
22: #endif
1.53 moko 23:
1.1 paf 24: // defines
25:
1.19 misha 26: #define HTTP_METHOD_NAME "method"
27: #define HTTP_FORM_NAME "form"
28: #define HTTP_BODY_NAME "body"
29: #define HTTP_TIMEOUT_NAME "timeout"
30: #define HTTP_HEADERS_NAME "headers"
1.22 misha 31: #define HTTP_FORM_ENCTYPE_NAME "enctype"
1.19 misha 32: #define HTTP_ANY_STATUS_NAME "any-status"
1.59 moko 33: #define HTTP_OMIT_POST_CHARSET_NAME "omit-post-charset" // ^file::load[...;http://...;$.method[post]] by default adds charset to content-type
1.12 misha 34:
1.1 paf 35: #define HTTP_USER "user"
36: #define HTTP_PASSWORD "password"
37:
1.70 moko 38: #define HTTP_USER_AGENT "user-agent"
1.1 paf 39: #define DEFAULT_USER_AGENT "parser3"
40:
1.59 moko 41: #ifndef INADDR_NONE
42: #define INADDR_NONE ((ulong) -1)
43: #endif
1.1 paf 44:
45: #undef CRLF
46: #define CRLF "\r\n"
47:
1.54 misha 48: // helpers
1.56 misha 49:
1.54 misha 50: class Cookies_table_template_columns: public ArrayString {
51: public:
52: Cookies_table_template_columns() {
53: *this+=new String("name");
54: *this+=new String("value");
55: *this+=new String("expires");
56: *this+=new String("max-age");
57: *this+=new String("domain");
58: *this+=new String("path");
59: *this+=new String("httponly");
60: *this+=new String("secure");
61: }
62: };
63:
64:
1.1 paf 65: static bool set_addr(struct sockaddr_in *addr, const char* host, const short port){
1.22 misha 66: memset(addr, 0, sizeof(*addr));
67: addr->sin_family=AF_INET;
68: addr->sin_port=htons(port);
69: if(host) {
1.65 moko 70: struct hostent *hostIP=gethostbyname(host);
71: if(hostIP && hostIP->h_addrtype == AF_INET){
72: memcpy(&addr->sin_addr, hostIP->h_addr, hostIP->h_length);
73: return true;
74: }
75: }
76: return false;
1.1 paf 77: }
78:
79: size_t guess_content_length(char* buf) {
80: char* ptr;
81: if((ptr=strstr(buf, "Content-Length:"))) // Apache
82: goto found;
1.37 misha 83: if((ptr=strstr(buf, "content-length:"))) // Parser 3 before 3.4.0
1.1 paf 84: goto found;
85: if((ptr=strstr(buf, "Content-length:"))) // maybe 1
86: goto found;
87: if((ptr=strstr(buf, "CONTENT-LENGTH:"))) // maybe 2
88: goto found;
89: return 0;
90: found:
91: char *error_pos;
1.37 misha 92: size_t result=(size_t)strtol(ptr+15/*strlen("Content-Length:")*/, &error_pos, 0);
1.1 paf 93:
94: const size_t reasonable_initial_max=0x400*0x400*10 /*10M*/;
95: if(result>reasonable_initial_max) // sanity check
96: return reasonable_initial_max;
97: return 0;//result;
98: }
99:
100: static int http_read_response(char*& response, size_t& response_size, int sock, bool fail_on_status_ne_200) {
101: int result=0;
1.37 misha 102: // fetching some to local buffer, guessing on possible Content-Length
103: response_size=0x400*20; // initial size if Content-Length could not be determined
1.1 paf 104: const size_t preview_size=0x400*20;
105: char preview_buf[preview_size+1/*terminator*/]; // 20K buffer to preview headers
106: ssize_t received_size=recv(sock, preview_buf, preview_size, 0);
107: if(received_size==0)
108: goto done;
109: if(received_size<0) {
110: if(int no=pa_socks_errno())
111: throw Exception("http.timeout",
112: 0,
113: "error receiving response header: %s (%d)", pa_socks_strerr(no), no);
114: goto done;
115: }
1.2 paf 116: // terminator [helps futher string searches]
117: preview_buf[received_size]=0;
118: // checking status
119: if(char* EOLat=strstr(preview_buf, "\n")) {
120: const String status_line(pa_strdup(preview_buf, EOLat-preview_buf));
121: ArrayString astatus;
122: size_t pos_after=0;
123: status_line.split(astatus, pos_after, " ");
124: const String& status_code=*astatus.get(astatus.count()>1?1:0);
125: result=status_code.as_int();
126:
127: if(fail_on_status_ne_200 && result!=200)
128: throw Exception("http.status",
129: &status_code,
130: "invalid HTTP response status");
131: }
1.1 paf 132: // detecting response_size
133: {
134: if(size_t content_length=guess_content_length(preview_buf))
135: response_size=preview_size+content_length; // a little more than needed, will adjust response_size by actual received size later
136: }
137:
138: // [gcc is happier this way, see goto above]
139: {
140: // allocating initial buf
141: response=(char*)pa_malloc_atomic(response_size+1/*terminator*/); // just setting memory block type
142: char* ptr=response;
143: size_t todo_size=response_size;
144: // coping part of already received body
145: memcpy(ptr, preview_buf, received_size);
146: ptr+=received_size;
147: todo_size-=received_size;
148:
149: // we use terminator byte for two purposes here:
150: // 1. we return there zero always, not knowing: maybe they would want to create String form $file.body?
151: // invariant: all Strings should have zero-terminated buffers
1.37 misha 152: // 2. we use that out-of-size byte to detect if our Content-Length guess was wrong
1.1 paf 153: // when recv gets more than we expected
1.37 misha 154: // a) we know that the Content-Length guess was wrong
1.1 paf 155: // b) we have space to put the first byte of extra data
156: // c) we use less code to detect normal situation: on last while-cycle recv expected to just return 0
157: while(true) {
158: received_size=recv(sock, ptr, todo_size+1/*there is always a place for terminator*/, 0);
159: if(received_size==0) {
160: response_size-=todo_size; // in case we received less than expected, cut down the reported size
161: break;
162: }
163: if(received_size<0) {
164: if(int no=pa_socks_errno())
165: throw Exception("http.timeout",
166: 0,
167: "error receiving response body: %s (%d)", pa_socks_strerr(no), no);
168: break;
169: }
170: // they've touched the terminator?
171: if((size_t)received_size>todo_size)
172: {
173: // that means that our guessed response_size was not big enough
174: const size_t grow_chunk_size=0x400*0x400; // 1M
175: response_size+=grow_chunk_size;
176: size_t ptr_offset=ptr-response;
177: response=(char*)pa_realloc(response, response_size+1/*terminator*/);
178: ptr=response+ptr_offset;
179: todo_size+=grow_chunk_size;
180: }
181: // can't do this before realloc: we need <todo_size check
182: ptr+=received_size;
183: todo_size-=received_size;
184: }
185: }
186: done:
187: if(result)
188: {
189: response[response_size]=0;
190: return result;
191: }
192: else
193: throw Exception("http.response",
194: 0,
195: "bad response from host - no status found (size=%u)", response_size);
196: }
197:
198: /* ********************** request *************************** */
199:
200: #if defined(SIGALRM) && defined(HAVE_SIGSETJMP) && defined(HAVE_SIGLONGJMP)
201: # define PA_USE_ALARM
202: #endif
203:
204: #ifdef PA_USE_ALARM
205: static sigjmp_buf timeout_env;
206: static void timeout_handler(int /*sig*/){
1.22 misha 207: siglongjmp(timeout_env, 1);
1.1 paf 208: }
209: #endif
210:
211: static int http_request(char*& response, size_t& response_size,
212: const char* host, short port,
1.22 misha 213: const char* request, size_t request_size,
1.1 paf 214: int timeout_secs,
215: bool fail_on_status_ne_200) {
216: if(!host)
217: throw Exception("http.host",
218: 0,
219: "zero hostname"); //never
220:
221: volatile // to prevent makeing it register variable, because it will be clobbered by longjmp [thanks gcc warning]
222: int sock=-1;
223: #ifdef PA_USE_ALARM
224: signal(SIGALRM, timeout_handler);
225: #endif
226: #ifdef PA_USE_ALARM
227: if(sigsetjmp(timeout_env, 1)) {
228: // stupid gcc [2.95.4] generated bad code
229: // which failed to handle sigsetjmp+throw: crashed inside of pre-throw code.
230: // rewritten simplier [athough duplicating closesocket code]
231: if(sock>=0)
232: closesocket(sock);
233: throw Exception("http.timeout",
234: 0,
235: "timeout occured while retrieving document");
236: return 0; // never
237: } else {
238: alarm(timeout_secs);
239: #endif
240: try {
241: int result;
242: struct sockaddr_in dest;
243:
244: if(!set_addr(&dest, host, port))
245: throw Exception("http.host",
246: 0,
247: "can not resolve hostname \"%s\"", host);
248:
249: if((sock=socket(AF_INET, SOCK_STREAM, IPPROTO_TCP/*0*/))<0) {
250: int no=pa_socks_errno();
251: throw Exception("http.connect",
252: 0,
253: "can not make socket: %s (%d)", pa_socks_strerr(no), no);
254: }
255:
256: // To enable SO_DONTLINGER (that is, disable SO_LINGER)
257: // l_onoff should be set to zero and setsockopt should be called
258: linger dont_linger={0,0};
259: setsockopt(sock, SOL_SOCKET, SO_LINGER, (const char *)&dont_linger, sizeof(dont_linger));
260:
261: #ifdef WIN32
262: // SO_*TIMEO can be defined in .h but not implemlemented in protocol,
263: // failing subsequently with Option not supported by protocol (99) message
264: // could not suppress that, so leaving this only for win32
265: int timeout_ms=timeout_secs*1000;
266: setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO, (const char*)&timeout_ms, sizeof(timeout_ms));
267: setsockopt(sock, SOL_SOCKET, SO_RCVTIMEO, (const char*)&timeout_ms, sizeof(timeout_ms));
268: #endif
269:
270: if(connect(sock, (struct sockaddr *)&dest, sizeof(dest))) {
271: int no=pa_socks_errno();
272: throw Exception("http.connect",
273: 0,
274: "can not connect to host \"%s\": %s (%d)", host, pa_socks_strerr(no), no);
275: }
1.22 misha 276:
1.1 paf 277: if(send(sock, request, request_size, 0)!=(ssize_t)request_size) {
278: int no=pa_socks_errno();
279: throw Exception("http.timeout",
280: 0,
281: "error sending request: %s (%d)", pa_socks_strerr(no), no);
282: }
283:
284: result=http_read_response(response, response_size, sock, fail_on_status_ne_200);
285: closesocket(sock);
286: #ifdef PA_USE_ALARM
287: alarm(0);
288: #endif
289: return result;
290: } catch(...) {
291: #ifdef PA_USE_ALARM
292: alarm(0);
293: #endif
294: if(sock>=0)
295: closesocket(sock);
296: rethrow;
297: }
298: #ifdef PA_USE_ALARM
299: }
300: #endif
301: }
302:
303: #ifndef DOXYGEN
304: struct Http_pass_header_info {
305: Request_charsets* charsets;
306: String* request;
1.35 misha 307: bool* user_agent_specified;
308: bool* content_type_specified;
309: bool* content_type_url_encoded;
1.1 paf 310: };
311: #endif
1.50 moko 312:
313: char *pa_http_safe_header_name(const char *name) {
314: char *result=pa_strdup(name);
315: char *n=result;
1.52 misha 316: if(!pa_isalpha((unsigned char)*n))
1.50 moko 317: *n++ = '_';
318: for(; *n; ++n) {
1.52 misha 319: if (!pa_isalnum((unsigned char)*n) && *n != '-' && *n != '_')
1.50 moko 320: *n = '_';
321: }
322: return result;
323: }
324:
1.35 misha 325: static void http_pass_header(HashStringValue::key_type aname,
326: HashStringValue::value_type avalue,
1.22 misha 327: Http_pass_header_info *info) {
1.9 misha 328:
1.41 misha 329: const char* name_cstr=aname.cstr();
330:
1.38 misha 331: if(strcasecmp(name_cstr, HTTP_CONTENT_LENGTH)==0)
332: return;
333:
1.50 moko 334: String name=String(pa_http_safe_header_name(capitalize(name_cstr)), String::L_AS_IS);
335: String value=attributed_meaning_to_string(*avalue, String::L_HTTP_HEADER, true);
1.9 misha 336:
1.35 misha 337: *info->request << name << ": " << value << CRLF;
1.1 paf 338:
1.38 misha 339: if(strcasecmp(name_cstr, HTTP_USER_AGENT)==0)
1.35 misha 340: *info->user_agent_specified=true;
1.38 misha 341: if(strcasecmp(name_cstr, HTTP_CONTENT_TYPE)==0){
1.35 misha 342: *info->content_type_specified=true;
1.62 moko 343: *info->content_type_url_encoded=pa_strncasecmp(value.cstr(), HTTP_CONTENT_TYPE_FORM_URLENCODED)==0;
1.35 misha 344: }
1.1 paf 345: }
346:
1.10 misha 347: static void http_pass_cookie(HashStringValue::key_type name,
1.20 misha 348: HashStringValue::value_type value,
349: Http_pass_header_info *info) {
1.10 misha 350:
1.17 misha 351: *info->request << String(name, String::L_HTTP_COOKIE) << "="
1.31 misha 352: << attributed_meaning_to_string(*value, String::L_HTTP_COOKIE, true)
1.10 misha 353: << "; ";
354:
355: }
1.1 paf 356:
357: static const String* basic_authorization_field(const char* user, const char* pass) {
358: if(!user&& !pass)
359: return 0;
360:
361: String combined;
362: if(user)
363: combined<<user;
364: combined<<":";
365: if(pass)
366: combined<<pass;
367:
1.20 misha 368: String* result=new String("Basic ");
369: *result<<pa_base64_encode(combined.cstr(), combined.length());
1.1 paf 370: return result;
371: }
372:
373: static void form_string_value2string(
1.20 misha 374: HashStringValue::key_type key,
375: const String& value,
376: String& result)
1.1 paf 377: {
1.30 misha 378: result << String(key, String::L_URI) << "=" << String(value, String::L_URI) << "&";
1.1 paf 379: }
1.20 misha 380:
1.1 paf 381: #ifndef DOXYGEN
382: struct Form_table_value2string_info {
383: HashStringValue::key_type key;
384: String& result;
385:
386: Form_table_value2string_info(HashStringValue::key_type akey, String& aresult):
387: key(akey), result(aresult) {}
388: };
389: #endif
390: static void form_table_value2string(Table::element_type row, Form_table_value2string_info* info) {
391: form_string_value2string(info->key, *row->get(0), info->result);
392: }
393: static void form_value2string(
1.20 misha 394: HashStringValue::key_type key,
395: HashStringValue::value_type value,
396: String* result)
1.1 paf 397: {
398: if(const String* svalue=value->get_string())
399: form_string_value2string(key, *svalue, *result);
400: else if(Table* tvalue=value->get_table()) {
401: Form_table_value2string_info info(key, *result);
402: tvalue->for_each(form_table_value2string, &info);
403: } else
1.18 misha 404: throw Exception(PARSER_RUNTIME,
1.1 paf 405: new String(key, String::L_TAINTED),
1.63 moko 406: "is %s, " HTTP_FORM_NAME " option value can be string or table only (file is allowed for $." HTTP_METHOD_NAME "[POST] + $." HTTP_FORM_ENCTYPE_NAME "[" HTTP_CONTENT_TYPE_MULTIPART_FORMDATA "])", value->type());
1.1 paf 407: }
1.20 misha 408:
1.5 misha 409: const char* pa_form2string(HashStringValue& form, Request_charsets& charsets) {
1.1 paf 410: String string;
1.3 paf 411: form.for_each<String*>(form_value2string, &string);
1.44 misha 412: return string.untaint_and_transcode_cstr(String::L_URI, &charsets);
1.1 paf 413: }
1.22 misha 414:
415: struct FormPart {
416: Request* r;
417: const char* boundary;
1.48 moko 418: String* string;
1.22 misha 419: Form_table_value2string_info* info;
1.48 moko 420:
421: struct BinaryBlock{
422: const char* ptr;
423: size_t length;
424:
425: BinaryBlock(String* astring, Request* r): ptr(astring->untaint_and_transcode_cstr(String::L_AS_IS, &r->charsets)), length(strlen(ptr)){}
426: BinaryBlock(const char* aptr, size_t alength): ptr(aptr), length(alength){}
427: };
428:
429: Array<BinaryBlock> blocks;
430:
431: FormPart(Request* ar, const char* aboundary): r(ar), boundary(aboundary), string(new String()){}
432:
433: const char *post(size_t &length){
434: if(blocks.count()){
435: blocks+=BinaryBlock(string, r);
436:
437: length=0;
438: for(size_t i=0; i<blocks.count(); i++)
439: length+=blocks[i].length;
440:
441: char *result=(char *)pa_malloc_atomic(length);
442: char *ptr=result;
443:
444: for(size_t i=0; i<blocks.count(); i++){
445: memcpy(ptr, blocks[i].ptr, blocks[i].length);
446: ptr+=blocks[i].length;
447: }
448:
449: return result;
450: } else {
451: BinaryBlock result(string, r);
452: length=result.length;
453: return result.ptr;
454: }
455: }
1.22 misha 456: };
457:
1.28 misha 458: static void form_part_boundary_header(FormPart& part, String::Body name, const char* file_name=0){
1.48 moko 459: *part.string << "--" << part.boundary
1.41 misha 460: << CRLF CONTENT_DISPOSITION_CAPITALIZED ": form-data; name=\""
1.48 moko 461: << name
1.28 misha 462: << "\"";
1.22 misha 463: if(file_name){
464: if(strcmp(file_name, NONAME_DAT)!=0)
1.48 moko 465: *part.string << "; filename=\"" << file_name << "\"";
466: *part.string << CRLF HTTP_CONTENT_TYPE_CAPITALIZED ": " << part.r->mime_type_of(file_name);
1.22 misha 467: }
1.48 moko 468: *part.string << CRLF CRLF;
1.22 misha 469: }
470:
471: static void form_string_value2part(
1.28 misha 472: HashStringValue::key_type key,
473: const String& value,
474: FormPart& part)
1.22 misha 475: {
1.28 misha 476: form_part_boundary_header(part, key);
1.48 moko 477: *part.string << value << CRLF;
1.22 misha 478: }
479:
480: static void form_file_value2part(
1.28 misha 481: HashStringValue::key_type key,
482: VFile& vfile,
483: FormPart& part)
1.22 misha 484: {
1.28 misha 485: form_part_boundary_header(part, key, vfile.fields().get(name_name)->as_string().cstr());
1.48 moko 486: part.blocks+=FormPart::BinaryBlock(part.string, part.r);
487: part.blocks+=FormPart::BinaryBlock(vfile.value_ptr(), vfile.value_size());
488: part.string=new String();
489: *part.string << CRLF;
1.22 misha 490: }
491:
492: static void form_table_value2part(Table::element_type row, FormPart* part) {
493: form_string_value2part(part->info->key, *row->get(0), *part);
494: }
495:
496: static void form_value2part(
1.28 misha 497: HashStringValue::key_type key,
498: HashStringValue::value_type value,
499: FormPart& part)
1.22 misha 500: {
501: if(const String* svalue=value->get_string())
502: form_string_value2part(key, *svalue, part);
503: else if(Table* tvalue=value->get_table()) {
1.48 moko 504: Form_table_value2string_info info(key, *part.string);
1.22 misha 505: part.info = &info;
506: tvalue->for_each(form_table_value2part, &part);
1.33 misha 507: } else if(VFile* vfile=static_cast<VFile *>(value->as("file"))){
1.22 misha 508: form_file_value2part(key, *vfile, part);
509: } else
510: throw Exception(PARSER_RUNTIME,
511: new String(key, String::L_TAINTED),
1.63 moko 512: "is %s, " HTTP_FORM_NAME " option value can be string, table or file only", value->type());
1.22 misha 513: }
514:
515: const char* pa_form2string_multipart(HashStringValue& form, Request& r, const char* boundary, size_t& post_size){
1.48 moko 516: FormPart formpart(&r, boundary);
1.22 misha 517: form.for_each<FormPart&>(form_value2part, formpart);
1.48 moko 518: *formpart.string << "--" << boundary << "--";
519: // @todo: return binary blocks here to save memory in pa_internal_file_read_http
520: return formpart.post(post_size);
1.22 misha 521: }
522:
1.1 paf 523: static void find_headers_end(char* p,
524: char*& headers_end_at,
525: char*& raw_body)
526: {
527: raw_body=p;
528: // \n\n
529: // \r\n\r\n
530: while((p=strchr(p, '\n'))) {
531: headers_end_at=++p; // \n>.<
532: if(*p=='\r') // \r\n>\r?<\n
533: p++;
534: if(*p=='\n') { // \r\n\r>\n?<
535: raw_body=p+1;
536: return;
537: }
538: }
539: headers_end_at=0;
540: }
541:
1.54 misha 542: // Set-Cookie: name=value; Domain=docs.foo.com; Path=/accounts; Expires=Wed, 13-Jan-2021 22:23:01 GMT; Secure; HttpOnly
543: static ArrayString* parse_cookie(Request& r, const String& cookie) {
1.64 moko 544: char *current=pa_strdup(cookie.cstr());
1.54 misha 545:
546: const String* name=0;
1.55 moko 547: const String* value=&String::Empty;
548: const String* expires=&String::Empty;
549: const String* max_age=&String::Empty;
550: const String* path=&String::Empty;
551: const String* domain=&String::Empty;
552: const String* httponly=&String::Empty;
553: const String* secure=&String::Empty;
1.54 misha 554:
555: bool first_pair=true;
556:
557: do {
558: if(char *meaning=search_stop(current, ';'))
559: if(char *attribute=search_stop(meaning, '=')) {
560: const String* sname=new String(unescape_chars(attribute, strlen(attribute), &r.charsets.source(), true/*don't convert '"' to space*/), String::L_TAINTED);
561: const String* smeaning=0;
562: if(meaning)
563: smeaning=new String(unescape_chars(meaning, strlen(meaning), &r.charsets.source(), true/*don't convert '"' to space*/), String::L_TAINTED);
564:
565: if(first_pair) {
566: // name + value
567: name=sname;
568: value=smeaning;
569: first_pair=false;
570: } else {
571: const String& slower=sname->change_case(r.charsets.source(), String::CC_LOWER);
572:
573: if(slower == "expires")
574: expires=smeaning;
575: else if(slower == "max-age")
576: max_age=smeaning;
577: else if(slower == "domain")
578: domain=smeaning;
579: else if(slower == "path")
580: path=smeaning;
581: else if(slower == "httponly")
582: httponly=new String("1", String::L_CLEAN);
583: else if(slower == "secure")
584: secure=new String("1", String::L_CLEAN);
585: else {
586: // todo@ ?
587: }
588: }
589: }
590: } while(current);
591:
592: if(!name)
593: return 0;
594:
595: ArrayString* result=new ArrayString(8);
596: *result+=name;
597: *result+=value;
598: *result+=expires;
599: *result+=max_age;
600: *result+=domain;
601: *result+=path;
602: *result+=httponly;
603: *result+=secure;
604:
605: return result;
606: }
607:
1.56 misha 608: Table* parse_cookies(Request& r, Table *cookies){
609: Table& result=*new Table(new Cookies_table_template_columns);
610:
611: for(Array_iterator<Table::element_type> i(*cookies); i.has_next(); )
612: if(ArrayString* row=parse_cookie(r, *i.next()->get(0)))
613: result+=row;
614:
615: return &result;
616: }
617:
1.72 ! moko 618: void *tables_update(HashStringValue& tables, const String::Body name, const String& value){
! 619: Table *table;
! 620: if(Value *valready=tables.get(name)) {
! 621: // second+ appearence
! 622: table=valready->get_table();
! 623: } else {
! 624: // first appearence
! 625: Table::columns_type columns=new ArrayString(1);
! 626: *columns+=new String("value");
! 627: table=new Table(columns);
! 628: tables.put(name, new VTable(table));
! 629: }
! 630: // this string becomes next row
! 631: ArrayString& row=*new ArrayString(1);
! 632: row+=&value;
! 633: *table+=&row;
! 634: }
! 635:
1.1 paf 636: /// @todo build .cookies field. use ^file.tables.SET-COOKIES.menu{ for now
1.72 ! moko 637: File_read_http_result pa_internal_file_read_http(Request& r, const String& file_spec, bool as_text, HashStringValue *options, bool transcode_text_result) {
1.1 paf 638: File_read_http_result result;
1.20 misha 639: char host[MAX_STRING];
1.66 moko 640: const char *idna_host;
1.1 paf 641: const char* uri;
1.49 moko 642: short port=80;
1.10 misha 643: const char* method="GET";
1.21 misha 644: bool method_is_get=true;
1.1 paf 645: HashStringValue* form=0;
646: int timeout_secs=2;
647: bool fail_on_status_ne_200=true;
1.12 misha 648: bool omit_post_charset=false;
1.1 paf 649: Value* vheaders=0;
1.10 misha 650: Value* vcookies=0;
1.11 misha 651: Value* vbody=0;
1.72 ! moko 652: Charset* asked_remote_charset=0;
1.58 moko 653: Charset* real_remote_charset=0;
1.1 paf 654: const char* user_cstr=0;
655: const char* password_cstr=0;
1.22 misha 656: const char* encode=0;
657: bool multipart=false;
1.1 paf 658:
659: if(options) {
660: int valid_options=pa_get_valid_file_options_count(*options);
661:
662: if(Value* vmethod=options->get(HTTP_METHOD_NAME)) {
663: valid_options++;
1.21 misha 664: method=vmethod->as_string().change_case(r.charsets.source(), String::CC_UPPER).cstr();
665: method_is_get=strcmp(method, "GET")==0;
1.1 paf 666: }
1.22 misha 667: if(Value* vencode=options->get(HTTP_FORM_ENCTYPE_NAME)) {
668: valid_options++;
669: encode=vencode->as_string().cstr();
670: }
1.1 paf 671: if(Value* vform=options->get(HTTP_FORM_NAME)) {
672: valid_options++;
673: form=vform->get_hash();
674: }
1.11 misha 675: if(vbody=options->get(HTTP_BODY_NAME)) {
1.1 paf 676: valid_options++;
677: }
678: if(Value* vtimeout=options->get(HTTP_TIMEOUT_NAME)) {
679: valid_options++;
680: timeout_secs=vtimeout->as_int();
681: }
1.11 misha 682: if(vheaders=options->get(HTTP_HEADERS_NAME)) {
1.1 paf 683: valid_options++;
684: }
1.11 misha 685: if(vcookies=options->get(HTTP_COOKIES_NAME)) {
1.10 misha 686: valid_options++;
687: }
1.1 paf 688: if(Value* vany_status=options->get(HTTP_ANY_STATUS_NAME)) {
689: valid_options++;
690: fail_on_status_ne_200=!vany_status->as_bool();
1.12 misha 691: }
1.20 misha 692: if(Value* vomit_post_charset=options->get(HTTP_OMIT_POST_CHARSET_NAME)){
1.12 misha 693: valid_options++;
694: omit_post_charset=vomit_post_charset->as_bool();
695: }
1.6 misha 696: if(Value* vcharset_name=options->get(PA_CHARSET_NAME)) {
1.58 moko 697: asked_remote_charset=&charsets.get(vcharset_name->as_string().change_case(r.charsets.source(), String::CC_UPPER));
698: }
699: if(Value* vresponse_charset_name=options->get(PA_RESPONSE_CHARSET_NAME)) {
1.61 moko 700: valid_options++;
1.58 moko 701: real_remote_charset=&charsets.get(vresponse_charset_name->as_string().change_case(r.charsets.source(), String::CC_UPPER));
1.1 paf 702: }
703: if(Value* vuser=options->get(HTTP_USER)) {
704: valid_options++;
705: user_cstr=vuser->as_string().cstr();
706: }
707: if(Value* vpassword=options->get(HTTP_PASSWORD)) {
708: valid_options++;
709: password_cstr=vpassword->as_string().cstr();
710: }
711:
712: if(valid_options!=options->count())
1.46 misha 713: throw Exception(PARSER_RUNTIME, 0, CALLED_WITH_INVALID_OPTION);
1.1 paf 714: }
715: if(!asked_remote_charset) // defaulting to $request:charset
1.22 misha 716: asked_remote_charset=&(r.charsets).source();
717:
718: if(encode){
719: if(method_is_get)
1.72 ! moko 720: throw Exception(PARSER_RUNTIME, 0, "you can not use $." HTTP_FORM_ENCTYPE_NAME " option with method GET");
1.22 misha 721:
722: multipart=strcasecmp(encode, HTTP_CONTENT_TYPE_MULTIPART_FORMDATA)==0;
723:
724: if(!multipart && strcasecmp(encode, HTTP_CONTENT_TYPE_FORM_URLENCODED)!=0)
1.72 ! moko 725: throw Exception(PARSER_RUNTIME, 0, "$." HTTP_FORM_ENCTYPE_NAME " option value can be " HTTP_CONTENT_TYPE_FORM_URLENCODED " or " HTTP_CONTENT_TYPE_MULTIPART_FORMDATA " only");
1.22 misha 726: }
1.1 paf 727:
1.11 misha 728: if(vbody){
729: if(method_is_get)
1.72 ! moko 730: throw Exception(PARSER_RUNTIME, 0, "you can not use $." HTTP_BODY_NAME " option with method GET");
1.11 misha 731:
732: if(form)
1.72 ! moko 733: throw Exception(PARSER_RUNTIME, 0, "you can not use options $." HTTP_BODY_NAME " and $." HTTP_FORM_NAME " together");
1.11 misha 734: }
1.1 paf 735:
736: //preparing request
1.29 misha 737: String& connect_string=*new String(file_spec);
1.1 paf 738:
1.48 moko 739: const char* request;
740: size_t request_size;
1.1 paf 741: {
742: // influence URLencoding of tainted pieces to String::L_URI lang
1.22 misha 743: Temp_client_charset temp(r.charsets, *asked_remote_charset);
1.1 paf 744:
1.44 misha 745: const char* connect_string_cstr=connect_string.untaint_and_transcode_cstr(String::L_URI, &(r.charsets));
1.1 paf 746:
747: const char* current=connect_string_cstr;
748: if(strncmp(current, "http://", 7)!=0)
1.72 ! moko 749: throw Exception(PARSER_RUNTIME, &connect_string, "does not start with http://"); //never
1.1 paf 750: current+=7;
751:
752: strncpy(host, current, sizeof(host)-1); host[sizeof(host)-1]=0;
1.34 misha 753: char* host_uri=lsplit(host, '/');
754: uri=host_uri?current+(host_uri-1-host):"/";
755: char* port_cstr=lsplit(host, ':');
1.49 moko 756:
757: if (port_cstr){
758: char* error_pos=0;
759: port=(short)strtol(port_cstr, &error_pos, 10);
760: if(port==0 || *error_pos)
761: throw Exception(PARSER_RUNTIME, &connect_string, "invalid port number '%s'", port_cstr);
762: }
1.1 paf 763:
1.66 moko 764: idna_host=pa_idna_encode(host, r.charsets.source());
765:
1.11 misha 766: // making request head
1.1 paf 767: String head;
1.11 misha 768: head << method << " " << uri;
1.28 misha 769: if(method_is_get && form)
770: head << (strchr(uri, '?')!=0?"&":"?") << pa_form2string(*form, r.charsets);
1.11 misha 771:
1.66 moko 772: head <<" HTTP/1.0" CRLF "Host: "<< idna_host;
1.49 moko 773: if (port != 80)
774: head << ":" << port_cstr;
775: head << CRLF;
1.11 misha 776:
1.71 moko 777: char* boundary= multipart ? get_uuid_boundary() : 0;
1.22 misha 778:
1.35 misha 779: String user_headers;
780: bool user_agent_specified=false;
781: bool content_type_specified=false;
782: bool content_type_url_encoded=false;
783: if(vheaders && !vheaders->is_string()) { // allow empty
784: if(HashStringValue *headers=vheaders->get_hash()) {
785: Http_pass_header_info info={
786: &(r.charsets),
787: &user_headers,
788: &user_agent_specified,
789: &content_type_specified,
790: &content_type_url_encoded};
791: headers->for_each<Http_pass_header_info*>(http_pass_header, &info);
792: } else
1.72 ! moko 793: throw Exception(PARSER_RUNTIME, 0, "headers param must be hash");
1.35 misha 794: };
795:
1.48 moko 796: const char* request_body=0;
1.22 misha 797: size_t post_size=0;
798: if(form && !method_is_get) {
1.38 misha 799: head << "Content-Type: " << (multipart ? HTTP_CONTENT_TYPE_MULTIPART_FORMDATA : HTTP_CONTENT_TYPE_FORM_URLENCODED);
1.28 misha 800:
801: if(!omit_post_charset)
802: head << "; charset=" << asked_remote_charset->NAME_CSTR();
803:
1.22 misha 804: if(multipart) {
1.28 misha 805: head << "; boundary=" << boundary;
1.48 moko 806: request_body=pa_form2string_multipart(*form, r/*charsets & mime_type needed*/, boundary, post_size/*correct post_size returned here*/);
1.22 misha 807: } else {
1.48 moko 808: request_body=pa_form2string(*form, r.charsets);
809: post_size=strlen(request_body);
1.22 misha 810: }
1.28 misha 811: head << CRLF;
1.35 misha 812: } else if(vbody) {
1.38 misha 813: // $.body was specified
1.35 misha 814: if(content_type_url_encoded){
1.36 misha 815: // transcode + url-encode
1.48 moko 816: request_body=vbody->as_string().untaint_and_transcode_cstr(String::L_URI, &(r.charsets));
1.35 misha 817: } else {
1.36 misha 818: // content-type != application/x-www-form-urlencoded -> transcode only, don't url-encode!
1.72 ! moko 819: const String &sbody=vbody->as_string();
! 820: request_body=Charset::transcode(String::C(sbody.cstr(), sbody.length()), r.charsets.source(), *asked_remote_charset).str;
1.35 misha 821: }
1.48 moko 822: post_size=strlen(request_body);
1.1 paf 823: }
824:
825: // http://www.ietf.org/rfc/rfc2617.txt
826: if(const String* authorization_field_value=basic_authorization_field(user_cstr, password_cstr))
1.38 misha 827: head << "Authorization: " << *authorization_field_value << CRLF;
1.1 paf 828:
1.35 misha 829: head << user_headers;
830:
1.1 paf 831: if(!user_agent_specified) // defaulting
1.38 misha 832: head << "User-Agent: " DEFAULT_USER_AGENT CRLF;
1.1 paf 833:
1.12 misha 834: if(form && !method_is_get && content_type_specified) // POST + form + content-type was specified
1.72 ! moko 835: throw Exception(PARSER_RUNTIME, 0, "$.content-type can't be specified with method POST");
1.12 misha 836:
1.11 misha 837: if(vcookies && !vcookies->is_string()){ // allow empty
1.10 misha 838: if(HashStringValue* cookies=vcookies->get_hash()) {
1.37 misha 839: head << "Cookie: ";
1.35 misha 840: Http_pass_header_info info={&(r.charsets), &head, 0, 0, 0};
1.10 misha 841: cookies->for_each<Http_pass_header_info*>(http_pass_cookie, &info);
842: head << CRLF;
843: } else
1.72 ! moko 844: throw Exception(PARSER_RUNTIME, 0, "cookies param must be hash");
1.10 misha 845: }
846:
1.48 moko 847: if(request_body)
1.38 misha 848: head << "Content-Length: " << format(post_size, "%u") << CRLF;
1.48 moko 849:
850: head << CRLF;
851:
852: const char *request_head=head.untaint_and_transcode_cstr(String::L_URI, &(r.charsets));
1.1 paf 853:
1.48 moko 854: if(request_body){
855: size_t head_size = strlen(request_head);
856: request_size=post_size + head_size;
857: char *ptr=(char *)pa_malloc_atomic(request_size);
858: memcpy(ptr, request_head, head_size);
859: memcpy(ptr+head_size, request_body, post_size);
860: request=ptr;
861: } else {
862: request_size=strlen(request_head);
863: request=request_head;
864: }
1.1 paf 865: }
866:
1.72 ! moko 867: char* response_str;
1.1 paf 868: size_t response_size;
1.22 misha 869:
1.28 misha 870: // sending request
1.72 ! moko 871: int status_code=http_request(response_str, response_size, idna_host, port, request, request_size, timeout_secs, fail_on_status_ne_200);
1.1 paf 872:
1.72 ! moko 873: // processing results
1.1 paf 874: char* raw_body; size_t raw_body_size;
875: char* headers_end_at;
1.72 ! moko 876: find_headers_end(response_str, headers_end_at, raw_body);
! 877: raw_body_size=response_size-(raw_body-response_str);
1.1 paf 878:
879: result.headers=new HashStringValue;
880: VHash* vtables=new VHash;
1.72 ! moko 881: result.headers->put("tables", vtables);
! 882:
! 883: ResponseHeaders response;
1.1 paf 884:
885: if(headers_end_at) {
886: *headers_end_at=0;
1.72 ! moko 887: const String header_block(String::C(response_str, headers_end_at-response_str), String::L_TAINTED);
1.1 paf 888:
889: ArrayString aheaders;
890:
891: size_t pos_after=0;
892: header_block.split(aheaders, pos_after, "\n");
893:
1.72 ! moko 894: Array_iterator<const String*> i(aheaders);
! 895: i.next(); // skipping status
! 896: for(;i.has_next();){
! 897: const char *line=i.next()->cstr();
! 898: if(!response.add_header(line))
! 899: throw Exception("http.response", &connect_string, "bad response from host - bad header \"%s\"", line);
1.1 paf 900: }
1.72 ! moko 901: }
1.54 misha 902:
1.72 ! moko 903: if (!real_remote_charset && !response.content_type.is_empty())
! 904: real_remote_charset= detect_charset(response.content_type.cstr());
1.1 paf 905:
1.72 ! moko 906: if(as_text)
1.68 moko 907: real_remote_charset=charsets.checkBOM(raw_body, raw_body_size, real_remote_charset);
1.72 ! moko 908:
! 909: if (!real_remote_charset)
! 910: real_remote_charset=asked_remote_charset; // never null
! 911:
! 912: for(Array_iterator<ResponseHeaders::Header> i(response.headers); i.has_next(); ){
! 913: ResponseHeaders::Header header=i.next();
! 914:
! 915: header.transcode(*real_remote_charset, r.charsets.source());
! 916:
! 917: String &header_value=*new String(header.value, String::L_TAINTED);
! 918:
! 919: tables_update(vtables->hash(), header.name, header_value);
! 920: result.headers->put(header.name, new VString(header_value));
1.16 misha 921: }
922:
1.72 ! moko 923: // filling $.cookies
! 924: if(Value *vcookies=vtables->hash().get("SET-COOKIE"))
! 925: result.headers->put(HTTP_COOKIES_NAME, new VTable(parse_cookies(r, vcookies->get_table())));
! 926:
1.1 paf 927: // output response
928: String::C real_body=String::C(raw_body, raw_body_size);
1.16 misha 929:
930: if(as_text && transcode_text_result && raw_body_size) { // raw_body_size must be checked because transcode returns CONST string in case length==0, which contradicts hacking few lines below
1.22 misha 931: real_body=Charset::transcode(real_body, *real_remote_charset, r.charsets.source());
1.1 paf 932: }
933:
934: result.str=const_cast<char *>(real_body.str); // hacking a little
935: result.length=real_body.length;
1.16 misha 936:
1.22 misha 937: if(as_text && result.length)
938: fix_line_breaks(result.str, result.length);
939:
1.1 paf 940: result.headers->put(file_status_name, new VInt(status_code));
1.16 misha 941:
1.1 paf 942: return result;
943: }
E-mail:
![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to pa_http.C CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [parser3project] / parser3 / src / main