[apple/icu.git] / icuSources / tools / genbrk / genbrk.cpp

/*
**********************************************************************
*   Copyright (C) 2002-2006, International Business Machines
*   Corporation and others.  All Rights Reserved.
**********************************************************************
*
* File genbrk.c
*/

//--------------------------------------------------------------------
//
//   Tool for generating RuleBasedBreakIterator data files (.brk files).
//   .brk files contain the precompiled rules for standard types
//   of iterators - word, line, sentence, etc.
//
//   Usage:  genbrk [options] -r rule-file.txt  -o output-file.brk
//
//       options:   -v         verbose
//                  -? or -h   help
//
//   The input rule file is a plain text file containing break rules
//    in the input format accepted by RuleBasedBreakIterators.  The
//    file can be encoded as utf-8, or utf-16 (either endian), or
//    in the default code page (platform dependent.).  utf encoded
//    files must include a BOM.
//
//--------------------------------------------------------------------

#include "unicode/utypes.h"
#include "unicode/ucnv.h"
#include "unicode/unistr.h"
#include "unicode/rbbi.h"
#include "unicode/uclean.h"
#include "unicode/udata.h"
#include "unicode/putil.h"

#include "uoptions.h"
#include "unewdata.h"
#include "ucmndata.h"
#include "rbbidata.h"
#include "cmemory.h"

#include <stdio.h>
#include <stdlib.h>
#include <string.h>

static char *progName;
static UOption options[]={
    UOPTION_HELP_H,             /* 0 */
    UOPTION_HELP_QUESTION_MARK, /* 1 */
    UOPTION_VERBOSE,            /* 2 */
    { "rules", NULL, NULL, NULL, 'r', UOPT_REQUIRES_ARG, 0 },   /* 3 */
    { "out",   NULL, NULL, NULL, 'o', UOPT_REQUIRES_ARG, 0 },   /* 4 */
    UOPTION_ICUDATADIR,         /* 5 */
    UOPTION_DESTDIR,            /* 6 */
    UOPTION_COPYRIGHT,          /* 7 */
};

void usageAndDie(int retCode) {
        printf("Usage: %s [-v] [-options] -r rule-file -o output-file\n", progName);
        printf("\tRead in break iteration rules text and write out the binary data\n"
            "options:\n"
            "\t-h or -? or --help  this usage text\n"
            "\t-V or --version     show a version message\n"
            "\t-c or --copyright   include a copyright notice\n"
            "\t-v or --verbose     turn on verbose output\n"
            "\t-i or --icudatadir  directory for locating any needed intermediate data files,\n"
            "\t                    followed by path, defaults to %s\n"
            "\t-d or --destdir     destination directory, followed by the path\n",
            u_getDataDirectory());
        exit (retCode);
}


#if UCONFIG_NO_BREAK_ITERATION

/* dummy UDataInfo cf. udata.h */
static UDataInfo dummyDataInfo = {
    sizeof(UDataInfo),
    0,

    U_IS_BIG_ENDIAN,
    U_CHARSET_FAMILY,
    U_SIZEOF_UCHAR,
    0,

    { 0, 0, 0, 0 },                 /* dummy dataFormat */
    { 0, 0, 0, 0 },                 /* dummy formatVersion */
    { 0, 0, 0, 0 }                  /* dummy dataVersion */
};

#else

//
//  Set up the ICU data header, defined in ucmndata.h
//
DataHeader dh ={
    {sizeof(DataHeader),           // Struct MappedData
        0xda,
        0x27},

    {                               // struct UDataInfo
        sizeof(UDataInfo),          //     size
        0,                          //     reserved
        U_IS_BIG_ENDIAN,
        U_CHARSET_FAMILY,
        U_SIZEOF_UCHAR,
        0,                          //     reserved

    { 0x42, 0x72, 0x6b, 0x20 },     //     dataFormat="Brk "
    { 0xff, 0, 0, 0 },              //     formatVersion.  Filled in later with values
                                    //      from the RBBI rule builder.  The  values declared
                                    //      here should never appear in any real RBBI data.
        { 4, 1, 0, 0 }              //   dataVersion (Unicode version)
    }};

#endif

//----------------------------------------------------------------------------
//
//  main      for genbrk
//
//----------------------------------------------------------------------------
int  main(int argc, char **argv) {
    UErrorCode  status = U_ZERO_ERROR;
    const char *ruleFileName;
    const char *outFileName;
    const char *outDir = NULL;
    const char *copyright = NULL;

    //
    // Pick up and check the command line arguments,
    //    using the standard ICU tool utils option handling.
    //
    U_MAIN_INIT_ARGS(argc, argv);
    progName = argv[0];
    argc=u_parseArgs(argc, argv, sizeof(options)/sizeof(options[0]), options);
    if(argc<0) {
        // Unrecognized option
        fprintf(stderr, "error in command line argument \"%s\"\n", argv[-argc]);
        usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
    }

    if(options[0].doesOccur || options[1].doesOccur) {
        //  -? or -h for help.
        usageAndDie(0);
    }

    if (!(options[3].doesOccur && options[4].doesOccur)) {
        fprintf(stderr, "rule file and output file must both be specified.\n");
        usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
    }
    ruleFileName = options[3].value;
    outFileName  = options[4].value;

    if (options[5].doesOccur) {
        u_setDataDirectory(options[5].value);
    }

    /* Initialize ICU */
    u_init(&status);
    if (U_FAILURE(status)) {
        fprintf(stderr, "%s: can not initialize ICU.  status = %s\n",
            argv[0], u_errorName(status));
        exit(1);
    }
    status = U_ZERO_ERROR;

    /* Combine the directory with the file name */
    if(options[6].doesOccur) {
        outDir = options[6].value;
    }
    if (options[7].doesOccur) {
        copyright = U_COPYRIGHT_STRING;
    }

#if UCONFIG_NO_BREAK_ITERATION

    UNewDataMemory *pData;
    char msg[1024];

    /* write message with just the name */
    sprintf(msg, "genbrk writes dummy %s because of UCONFIG_NO_BREAK_ITERATION, see uconfig.h", outFileName);
    fprintf(stderr, "%s\n", msg);

    /* write the dummy data file */
    pData = udata_create(outDir, NULL, outFileName, &dummyDataInfo, NULL, &status);
    udata_writeBlock(pData, msg, strlen(msg));
    udata_finish(pData, &status);
    return (int)status;

#else

    //
    //  Read in the rule source file
    //
    long        result;
    long        ruleFileSize;
    FILE        *file;
    char        *ruleBufferC;

    file = fopen(ruleFileName, "rb");
    if( file == 0 ) {
        fprintf(stderr, "Could not open file \"%s\"\n", ruleFileName);
        exit(-1);
    }
    fseek(file, 0, SEEK_END);
    ruleFileSize = ftell(file);
    fseek(file, 0, SEEK_SET);
    ruleBufferC = new char[ruleFileSize+10];

    result = (long)fread(ruleBufferC, 1, ruleFileSize, file);
    if (result != ruleFileSize)  {
        fprintf(stderr, "Error reading file \"%s\"\n", ruleFileName);
        exit (-1);
    }
    ruleBufferC[ruleFileSize]=0;
    fclose(file);

    //
    // Look for a Unicode Signature (BOM) on the rule file
    //
    int32_t        signatureLength;
    const char *   ruleSourceC = ruleBufferC;
    const char*    encoding = ucnv_detectUnicodeSignature(
                           ruleSourceC, ruleFileSize, &signatureLength, &status);
    if (U_FAILURE(status)) {
        exit(status);
    }
    if(encoding!=NULL ){
        ruleSourceC  += signatureLength;
        ruleFileSize -= signatureLength;
    }

    //
    // Open a converter to take the rule file to UTF-16
    //
    UConverter* conv;
    conv = ucnv_open(encoding, &status);
    if (U_FAILURE(status)) {
        fprintf(stderr, "ucnv_open: ICU Error \"%s\"\n", u_errorName(status));
        exit(status);
    }

    //
    // Convert the rules to UChar.
    //  Preflight first to determine required buffer size.
    //
    uint32_t destCap = ucnv_toUChars(conv,
                       NULL,           //  dest,
                       0,              //  destCapacity,
                       ruleSourceC,
                       ruleFileSize,
                       &status);
    if (status != U_BUFFER_OVERFLOW_ERROR) {
        fprintf(stderr, "ucnv_toUChars: ICU Error \"%s\"\n", u_errorName(status));
        exit(status);
    };

    status = U_ZERO_ERROR;
    UChar *ruleSourceU = new UChar[destCap+1];
    ucnv_toUChars(conv,
                  ruleSourceU,     //  dest,
                  destCap+1,
                  ruleSourceC,
                  ruleFileSize,
                  &status);
    if (U_FAILURE(status)) {
        fprintf(stderr, "ucnv_toUChars: ICU Error \"%s\"\n", u_errorName(status));
        exit(status);
    };
    ucnv_close(conv);


    //
    //  Put the source rules into a UnicodeString
    //
    UnicodeString ruleSourceS(FALSE, ruleSourceU, destCap);

    //
    //  Create the break iterator from the rules
    //     This will compile the rules.
    //
    UParseError parseError;
    parseError.line = 0;
    parseError.offset = 0;
    RuleBasedBreakIterator *bi = new RuleBasedBreakIterator(ruleSourceS, parseError, status);
    if (U_FAILURE(status)) {
        fprintf(stderr, "createRuleBasedBreakIterator: ICU Error \"%s\"  at line %d, column %d\n",
                u_errorName(status), (int)parseError.line, (int)parseError.offset);
        exit(status);
    };


    //
    //  Get the compiled rule data from the break iterator.
    //
    uint32_t        outDataSize;
    const uint8_t  *outData;
    outData = bi->getBinaryRules(outDataSize);

    // Copy the data format version numbers from the RBBI data header into the UDataMemory header.
    uprv_memcpy(dh.info.formatVersion, ((RBBIDataHeader *)outData)->fFormatVersion, sizeof(dh.info.formatVersion));

    //
    //  Create the output file
    //
    size_t bytesWritten;
    UNewDataMemory *pData;
    pData = udata_create(outDir, NULL, outFileName, &(dh.info), copyright, &status);
    if(U_FAILURE(status)) {
        fprintf(stderr, "genbrk: Could not open output file \"%s\", \"%s\"\n", 
                         outFileName, u_errorName(status));
        exit(status);
    }


    //  Write the data itself.
    udata_writeBlock(pData, outData, outDataSize);
    // finish up 
    bytesWritten = udata_finish(pData, &status);
    if(U_FAILURE(status)) {
        fprintf(stderr, "genbrk: error %d writing the output file\n", status);
        exit(status);
    }
    
    if (bytesWritten != outDataSize) {
        fprintf(stderr, "Error writing to output file \"%s\"\n", outFileName);
        exit(-1);
    }

    delete bi;
    delete[] ruleSourceU;
    delete[] ruleBufferC;
    u_cleanup();


    printf("genbrk: tool completed successfully.\n");
    return 0;

#endif /* #if !UCONFIG_NO_BREAK_ITERATION */
}
Commit	Line	Data
b75a7d8f A	1	/*
b75a7d8f A	2	**********************************************************************
73c04bcf	3	* Copyright (C) 2002-2006, International Business Machines
b75a7d8f A	4	* Corporation and others. All Rights Reserved.
	5	**********************************************************************
	6	*
	7	* File genbrk.c
	8	*/
	9
	10	//--------------------------------------------------------------------
	11	//
	12	// Tool for generating RuleBasedBreakIterator data files (.brk files).
	13	// .brk files contain the precompiled rules for standard types
	14	// of iterators - word, line, sentence, etc.
	15	//
	16	// Usage: genbrk [options] -r rule-file.txt -o output-file.brk
	17	//
	18	// options: -v verbose
	19	// -? or -h help
	20	//
	21	// The input rule file is a plain text file containing break rules
	22	// in the input format accepted by RuleBasedBreakIterators. The
	23	// file can be encoded as utf-8, or utf-16 (either endian), or
	24	// in the default code page (platform dependent.). utf encoded
	25	// files must include a BOM.
	26	//
	27	//--------------------------------------------------------------------
	28
b75a7d8f A	29	#include "unicode/utypes.h"
	30	#include "unicode/ucnv.h"
	31	#include "unicode/unistr.h"
	32	#include "unicode/rbbi.h"
	33	#include "unicode/uclean.h"
	34	#include "unicode/udata.h"
374ca955	35	#include "unicode/putil.h"
b75a7d8f A	36
	37	#include "uoptions.h"
	38	#include "unewdata.h"
	39	#include "ucmndata.h"
73c04bcf A	40	#include "rbbidata.h"
73c04bcf A	41	#include "cmemory.h"
b75a7d8f A	42
	43	#include <stdio.h>
	44	#include <stdlib.h>
	45	#include <string.h>
	46
b75a7d8f A	47	static char *progName;
	48	static UOption options[]={
	49	UOPTION_HELP_H, /* 0 */
	50	UOPTION_HELP_QUESTION_MARK, /* 1 */
	51	UOPTION_VERBOSE, /* 2 */
	52	{ "rules", NULL, NULL, NULL, 'r', UOPT_REQUIRES_ARG, 0 }, /* 3 */
	53	{ "out", NULL, NULL, NULL, 'o', UOPT_REQUIRES_ARG, 0 }, /* 4 */
	54	UOPTION_ICUDATADIR, /* 5 */
374ca955 A	55	UOPTION_DESTDIR, /* 6 */
374ca955 A	56	UOPTION_COPYRIGHT, /* 7 */
b75a7d8f A	57	};
	58
	59	void usageAndDie(int retCode) {
374ca955 A	60	printf("Usage: %s [-v] [-options] -r rule-file -o output-file\n", progName);
	61	printf("\tRead in break iteration rules text and write out the binary data\n"
	62	"options:\n"
	63	"\t-h or -? or --help this usage text\n"
	64	"\t-V or --version show a version message\n"
	65	"\t-c or --copyright include a copyright notice\n"
	66	"\t-v or --verbose turn on verbose output\n"
	67	"\t-i or --icudatadir directory for locating any needed intermediate data files,\n"
	68	"\t followed by path, defaults to %s\n"
	69	"\t-d or --destdir destination directory, followed by the path\n",
	70	u_getDataDirectory());
b75a7d8f A	71	exit (retCode);
	72	}
	73
	74
	75	#if UCONFIG_NO_BREAK_ITERATION
	76
	77	/* dummy UDataInfo cf. udata.h */
	78	static UDataInfo dummyDataInfo = {
	79	sizeof(UDataInfo),
	80	0,
	81
	82	U_IS_BIG_ENDIAN,
	83	U_CHARSET_FAMILY,
	84	U_SIZEOF_UCHAR,
	85	0,
	86
	87	{ 0, 0, 0, 0 }, /* dummy dataFormat */
	88	{ 0, 0, 0, 0 }, /* dummy formatVersion */
	89	{ 0, 0, 0, 0 } /* dummy dataVersion */
	90	};
	91
	92	#else
	93
	94	//
	95	// Set up the ICU data header, defined in ucmndata.h
	96	//
	97	DataHeader dh ={
	98	{sizeof(DataHeader), // Struct MappedData
	99	0xda,
	100	0x27},
	101
	102	{ // struct UDataInfo
	103	sizeof(UDataInfo), // size
	104	0, // reserved
	105	U_IS_BIG_ENDIAN,
	106	U_CHARSET_FAMILY,
	107	U_SIZEOF_UCHAR,
	108	0, // reserved
	109
	110	{ 0x42, 0x72, 0x6b, 0x20 }, // dataFormat="Brk "
73c04bcf A	111	{ 0xff, 0, 0, 0 }, // formatVersion. Filled in later with values
	112	// from the RBBI rule builder. The values declared
	113	// here should never appear in any real RBBI data.
	114	{ 4, 1, 0, 0 } // dataVersion (Unicode version)
b75a7d8f A	115	}};
	116
	117	#endif
	118
	119	//----------------------------------------------------------------------------
	120	//
	121	// main for genbrk
	122	//
	123	//----------------------------------------------------------------------------
	124	int main(int argc, char **argv) {
	125	UErrorCode status = U_ZERO_ERROR;
	126	const char *ruleFileName;
	127	const char *outFileName;
	128	const char *outDir = NULL;
374ca955	129	const char *copyright = NULL;
b75a7d8f A	130
	131	//
	132	// Pick up and check the command line arguments,
	133	// using the standard ICU tool utils option handling.
	134	//
	135	U_MAIN_INIT_ARGS(argc, argv);
	136	progName = argv[0];
	137	argc=u_parseArgs(argc, argv, sizeof(options)/sizeof(options[0]), options);
	138	if(argc<0) {
	139	// Unrecognized option
	140	fprintf(stderr, "error in command line argument \"%s\"\n", argv[-argc]);
	141	usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
	142	}
	143
	144	if(options[0].doesOccur \|\| options[1].doesOccur) {
	145	// -? or -h for help.
	146	usageAndDie(0);
	147	}
	148
	149	if (!(options[3].doesOccur && options[4].doesOccur)) {
	150	fprintf(stderr, "rule file and output file must both be specified.\n");
	151	usageAndDie(U_ILLEGAL_ARGUMENT_ERROR);
	152	}
	153	ruleFileName = options[3].value;
	154	outFileName = options[4].value;
b75a7d8f A	155
	156	if (options[5].doesOccur) {
	157	u_setDataDirectory(options[5].value);
	158	}
	159
374ca955 A	160	/* Initialize ICU */
	161	u_init(&status);
	162	if (U_FAILURE(status)) {
	163	fprintf(stderr, "%s: can not initialize ICU. status = %s\n",
	164	argv[0], u_errorName(status));
	165	exit(1);
	166	}
	167	status = U_ZERO_ERROR;
	168
b75a7d8f A	169	/* Combine the directory with the file name */
	170	if(options[6].doesOccur) {
	171	outDir = options[6].value;
b75a7d8f	172	}
374ca955 A	173	if (options[7].doesOccur) {
374ca955 A	174	copyright = U_COPYRIGHT_STRING;
b75a7d8f	175	}
b75a7d8f A	176
	177	#if UCONFIG_NO_BREAK_ITERATION
	178
	179	UNewDataMemory *pData;
73c04bcf	180	char msg[1024];
b75a7d8f A	181
b75a7d8f A	182	/* write message with just the name */
73c04bcf	183	sprintf(msg, "genbrk writes dummy %s because of UCONFIG_NO_BREAK_ITERATION, see uconfig.h", outFileName);
b75a7d8f A	184	fprintf(stderr, "%s\n", msg);
b75a7d8f A	185
b75a7d8f	186	/* write the dummy data file */
73c04bcf	187	pData = udata_create(outDir, NULL, outFileName, &dummyDataInfo, NULL, &status);
b75a7d8f A	188	udata_writeBlock(pData, msg, strlen(msg));
	189	udata_finish(pData, &status);
	190	return (int)status;
	191
	192	#else
	193
	194	//
	195	// Read in the rule source file
	196	//
374ca955	197	long result;
b75a7d8f A	198	long ruleFileSize;
	199	FILE *file;
	200	char *ruleBufferC;
	201
	202	file = fopen(ruleFileName, "rb");
	203	if( file == 0 ) {
	204	fprintf(stderr, "Could not open file \"%s\"\n", ruleFileName);
	205	exit(-1);
	206	}
	207	fseek(file, 0, SEEK_END);
	208	ruleFileSize = ftell(file);
	209	fseek(file, 0, SEEK_SET);
	210	ruleBufferC = new char[ruleFileSize+10];
	211
374ca955	212	result = (long)fread(ruleBufferC, 1, ruleFileSize, file);
b75a7d8f A	213	if (result != ruleFileSize) {
	214	fprintf(stderr, "Error reading file \"%s\"\n", ruleFileName);
	215	exit (-1);
	216	}
	217	ruleBufferC[ruleFileSize]=0;
	218	fclose(file);
	219
	220	//
	221	// Look for a Unicode Signature (BOM) on the rule file
	222	//
	223	int32_t signatureLength;
	224	const char * ruleSourceC = ruleBufferC;
	225	const char* encoding = ucnv_detectUnicodeSignature(
	226	ruleSourceC, ruleFileSize, &signatureLength, &status);
	227	if (U_FAILURE(status)) {
	228	exit(status);
	229	}
	230	if(encoding!=NULL ){
	231	ruleSourceC += signatureLength;
	232	ruleFileSize -= signatureLength;
	233	}
	234
	235	//
	236	// Open a converter to take the rule file to UTF-16
	237	//
	238	UConverter* conv;
	239	conv = ucnv_open(encoding, &status);
	240	if (U_FAILURE(status)) {
	241	fprintf(stderr, "ucnv_open: ICU Error \"%s\"\n", u_errorName(status));
	242	exit(status);
	243	}
	244
	245	//
	246	// Convert the rules to UChar.
	247	// Preflight first to determine required buffer size.
	248	//
	249	uint32_t destCap = ucnv_toUChars(conv,
	250	NULL, // dest,
	251	0, // destCapacity,
	252	ruleSourceC,
	253	ruleFileSize,
	254	&status);
	255	if (status != U_BUFFER_OVERFLOW_ERROR) {
	256	fprintf(stderr, "ucnv_toUChars: ICU Error \"%s\"\n", u_errorName(status));
	257	exit(status);
	258	};
	259
	260	status = U_ZERO_ERROR;
	261	UChar *ruleSourceU = new UChar[destCap+1];
	262	ucnv_toUChars(conv,
	263	ruleSourceU, // dest,
	264	destCap+1,
	265	ruleSourceC,
	266	ruleFileSize,
	267	&status);
	268	if (U_FAILURE(status)) {
	269	fprintf(stderr, "ucnv_toUChars: ICU Error \"%s\"\n", u_errorName(status));
	270	exit(status);
	271	};
	272	ucnv_close(conv);
	273
	274
	275	//
	276	// Put the source rules into a UnicodeString
277	//
278	UnicodeString ruleSourceS(FALSE, ruleSourceU, destCap);
279
280	//
281	// Create the break iterator from the rules
282	// This will compile the rules.
283	//
284	UParseError parseError;
374ca955 A	285	parseError.line = 0;
374ca955 A	286	parseError.offset = 0;
b75a7d8f A	287	RuleBasedBreakIterator *bi = new RuleBasedBreakIterator(ruleSourceS, parseError, status);
	288	if (U_FAILURE(status)) {
	289	fprintf(stderr, "createRuleBasedBreakIterator: ICU Error \"%s\" at line %d, column %d\n",
374ca955	290	u_errorName(status), (int)parseError.line, (int)parseError.offset);
b75a7d8f A	291	exit(status);
	292	};
	293
	294
	295	//
	296	// Get the compiled rule data from the break iterator.
	297	//
	298	uint32_t outDataSize;
	299	const uint8_t *outData;
	300	outData = bi->getBinaryRules(outDataSize);
	301
73c04bcf A	302	// Copy the data format version numbers from the RBBI data header into the UDataMemory header.
73c04bcf A	303	uprv_memcpy(dh.info.formatVersion, ((RBBIDataHeader *)outData)->fFormatVersion, sizeof(dh.info.formatVersion));
b75a7d8f A	304
	305	//
	306	// Create the output file
	307	//
	308	size_t bytesWritten;
374ca955 A	309	UNewDataMemory *pData;
	310	pData = udata_create(outDir, NULL, outFileName, &(dh.info), copyright, &status);
	311	if(U_FAILURE(status)) {
	312	fprintf(stderr, "genbrk: Could not open output file \"%s\", \"%s\"\n",
	313	outFileName, u_errorName(status));
	314	exit(status);
b75a7d8f	315	}
73c04bcf A	316
73c04bcf A	317
b75a7d8f	318	// Write the data itself.
374ca955 A	319	udata_writeBlock(pData, outData, outDataSize);
	320	// finish up
	321	bytesWritten = udata_finish(pData, &status);
	322	if(U_FAILURE(status)) {
	323	fprintf(stderr, "genbrk: error %d writing the output file\n", status);
	324	exit(status);
	325	}
	326
b75a7d8f	327	if (bytesWritten != outDataSize) {
374ca955	328	fprintf(stderr, "Error writing to output file \"%s\"\n", outFileName);
b75a7d8f A	329	exit(-1);
	330	}
	331
b75a7d8f A	332	delete bi;
	333	delete[] ruleSourceU;
	334	delete[] ruleBufferC;
b75a7d8f A	335	u_cleanup();
	336
	337
	338	printf("genbrk: tool completed successfully.\n");
	339	return 0;
	340
	341	#endif /* #if !UCONFIG_NO_BREAK_ITERATION */
	342	}
374ca955	343