X-Git-Url: https://git.dlugolecki.net.pl/?a=blobdiff_plain;f=gedcom%2Fmultilex.c;h=e4128174d5e6393ea497a8b6faa9f0be079a44c2;hb=31fb8432ef01f9d5a08cc801a785a07b4e9d12bd;hp=9a57b628d2dc3cdf75a2f6a960c78465f66f4e25;hpb=eb3312e2460a43abb5936cd35f29d17001367d3b;p=gedcom-parse.git diff --git a/gedcom/multilex.c b/gedcom/multilex.c index 9a57b62..e412817 100644 --- a/gedcom/multilex.c +++ b/gedcom/multilex.c @@ -24,6 +24,7 @@ #include "gedcom_internal.h" #include "multilex.h" #include "encoding.h" +#include "encoding_state.h" #include "xref.h" int line_no = 0; @@ -33,24 +34,24 @@ lex_func lf; #define NEW_MODEL_FILE "new.ged" -int lexer_init(ENCODING enc, FILE* f) +int lexer_init(Encoding enc, FILE* f) { if (enc == ONE_BYTE) { lf = &gedcom_1byte_lex; gedcom_1byte_myinit(f); - set_encoding_width(enc); + set_read_encoding_width(enc); return open_conv_to_internal("ASCII"); } else if (enc == TWO_BYTE_HILO) { lf = &gedcom_hilo_lex; gedcom_hilo_myinit(f); - set_encoding_width(enc); + set_read_encoding_width(enc); return open_conv_to_internal("UNICODE"); } else if (enc == TWO_BYTE_LOHI) { lf = &gedcom_lohi_lex; gedcom_lohi_myinit(f); - set_encoding_width(enc); + set_read_encoding_width(enc); return open_conv_to_internal("UNICODE"); } else { @@ -68,50 +69,68 @@ int gedcom_lex() return (*lf)(); } +void rewind_file(FILE* f) +{ + if (fseek(f, 0, 0) != 0) + gedcom_warning(_("Error positioning input file: %s"), strerror(errno)); +} + int determine_encoding(FILE* f) { char first[2]; int read; + set_read_encoding_bom(WITHOUT_BOM); read = fread(first, 1, 2, f); if (read != 2) { gedcom_warning(_("Error reading from input file: %s"), strerror(errno)); + rewind_file(f); return ONE_BYTE; } else if ((first[0] == '0') && (first[1] == ' ')) { - gedcom_debug_print(_("One-byte encoding")); - if (fseek(f, 0, 0) != 0) - gedcom_warning(_("Error positioning input file: %s"), strerror(errno)); + gedcom_debug_print("One-byte encoding"); + rewind_file(f); return ONE_BYTE; } - else if ((first[0] == '\0') && (first[1] == '0')) - { - gedcom_debug_print(_("Two-byte encoding, high-low")); - if (fseek(f, 0, 0) != 0) - gedcom_warning(_("Error positioning input file: %s"), strerror(errno)); + else if ((first[0] == '\0') && (first[1] == '0')) { + gedcom_debug_print("Two-byte encoding, high-low"); + rewind_file(f); return TWO_BYTE_HILO; } - else if ((first[0] == '\xFE') && (first[1] == '\xFF')) - { - gedcom_debug_print(_("Two-byte encoding, high-low, with BOM")); + else if ((first[0] == '\xFE') && (first[1] == '\xFF')) { + gedcom_debug_print("Two-byte encoding, high-low, with BOM"); + set_read_encoding_bom(WITH_BOM); return TWO_BYTE_HILO; } - else if ((first[0] == '0') && (first[1] == '\0')) - { - gedcom_debug_print(_("Two-byte encoding, low-high")); - if (fseek(f, 0, 0) != 0) - gedcom_warning(_("Error positioning input file: %s"), strerror(errno)); + else if ((first[0] == '0') && (first[1] == '\0')) { + gedcom_debug_print("Two-byte encoding, low-high"); + rewind_file(f); return TWO_BYTE_LOHI; } - else if ((first[0] == '\xFF') && (first[1] == '\xFE')) - { - gedcom_debug_print(_("Two-byte encoding, low-high, with BOM")); + else if ((first[0] == '\xFF') && (first[1] == '\xFE')) { + gedcom_debug_print("Two-byte encoding, low-high, with BOM"); + set_read_encoding_bom(WITH_BOM); return TWO_BYTE_LOHI; } + else if ((first[0] == '\xEF') && (first[1] == '\xBB')) { + read = fread(first, 1, 1, f); + if (read != 1) { + gedcom_warning(_("Error reading from input file: %s"), strerror(errno)); + rewind_file(f); + } + else if (first[0] == '\xBF') { + set_read_encoding_bom(WITH_BOM); + gedcom_debug_print("UTF-8 encoding, with BOM"); + } + else { + gedcom_warning(_("Unknown encoding, falling back to one-byte")); + rewind_file(f); + } + return ONE_BYTE; + } else { gedcom_warning(_("Unknown encoding, falling back to one-byte")); - if (fseek(f, 0, 0) != 0) - gedcom_warning(_("Error positioning input file: %s"), strerror(errno)); + rewind_file(f); return ONE_BYTE; } } @@ -122,6 +141,7 @@ int gedcom_init() { init_called = 1; update_gconv_search_path(); + init_encodings(); if (!setlocale(LC_ALL, "") || ! bindtextdomain(PACKAGE, LOCALEDIR) || ! bind_textdomain_codeset(PACKAGE, INTERNAL_ENCODING)) @@ -132,7 +152,7 @@ int gedcom_init() int gedcom_parse_file(const char* file_name) { - ENCODING enc; + Encoding enc; int result = 1; FILE* file; @@ -146,7 +166,7 @@ int gedcom_parse_file(const char* file_name) file_name, strerror(errno)); } else { - init_encodings(); + line_no = 1; enc = determine_encoding(file); if (lexer_init(enc, file)) {