static void
pgf_print_cnccat(GuMapItor* fn, const void* key, void* value,
GuExn* err)
{
PgfPrintFn* clo = (PgfPrintFn*) fn;
PgfCId name = *((PgfCId *) key);
PgfCncCat* cnccat = *((PgfCncCat**) value);
GuWriter *wtr = clo->wtr;
gu_puts(" ", wtr, err);
gu_string_write(name, wtr, err);
gu_puts(" :=\n", wtr, err);
PgfCCat *start = gu_list_index(cnccat->cats, 0);
PgfCCat *end = gu_list_index(cnccat->cats, gu_list_length(cnccat->cats)-1);
gu_printf(wtr, err, " range [C%d..C%d]\n", start->fid, end->fid);
gu_puts(" labels [", wtr, err);
for (size_t i = 0; i < cnccat->n_lins; i++) {
if (i > 0) {
gu_puts("\n ", wtr, err);
}
gu_string_write(cnccat->labels[i], wtr, err);
}
gu_puts("]\n", wtr, err);
}
void
pgf_print_symbol(PgfSymbol sym, GuWriter *wtr, GuExn *err)
{
switch (gu_variant_tag(sym)) {
case PGF_SYMBOL_CAT: {
PgfSymbolCat* scat = gu_variant_data(sym);
gu_printf(wtr, err, "<%d,%d>", scat->d, scat->r);
break;
}
case PGF_SYMBOL_KS: {
PgfSymbolKS* sks = gu_variant_data(sym);
pgf_print_tokens(sks->tokens, wtr, err);
break;
}
case PGF_SYMBOL_KP: {
PgfSymbolKP* skp = gu_variant_data(sym);
gu_puts("pre {", wtr, err);
pgf_print_tokens(skp->default_form, wtr, err);
for (size_t i = 0; i < skp->n_forms; i++) {
gu_puts("; ", wtr, err);
pgf_print_tokens(skp->forms[i].form, wtr, err);
gu_puts(" / ", wtr, err);
size_t n_prefixes = gu_list_length(skp->forms[i].prefixes);
for (size_t j = 0; j < n_prefixes; j++) {
if (j > 0) gu_putc(' ', wtr, err);
GuString prefix = gu_list_index(skp->forms[i].prefixes, j);
gu_putc('"', wtr, err);
gu_string_write(prefix, wtr, err);
gu_putc('"', wtr, err);
}
}
gu_puts("}", wtr, err);
break;
}
case PGF_SYMBOL_LIT: {
PgfSymbolLit *slit = gu_variant_data(sym);
gu_printf(wtr, err, "{%d,%d}", slit->d, slit->r);
break;
}
case PGF_SYMBOL_VAR: {
PgfSymbolVar *svar = gu_variant_data(sym);
gu_printf(wtr, err, "<%d,$%d>", svar->d, svar->r);
break;
}
default:
gu_impossible();
}
}
static void
pgf_print_tokens(PgfTokens tokens, GuWriter *wtr, GuExn *err)
{
gu_putc('"', wtr, err);
size_t n_toks = gu_seq_length(tokens);
for (size_t i = 0; i < n_toks; i++) {
if (i > 0) gu_putc(' ', wtr, err);
PgfToken tok = gu_seq_get(tokens, PgfToken, i);
gu_string_write(tok, wtr, err);
}
gu_putc('"', wtr, err);
}
void
pgf_print_flag(GuMapItor* fn, const void* key, void* value,
GuExn* err)
{
PgfPrintFn* clo = (PgfPrintFn*) fn;
PgfCId flag = *((PgfCId *) key);
PgfLiteral lit = *((PgfLiteral *) value);
GuWriter *wtr = clo->wtr;
gu_puts(" flag ", wtr, err);
gu_string_write(flag, wtr, err);
gu_puts(" = ", wtr, err);
pgf_print_literal(lit, wtr, err);
gu_puts(";\n", wtr, err);
}
void
pgf_print_absfun(GuMapItor* fn, const void* key, void* value,
GuExn* err)
{
PgfPrintFn* clo = (PgfPrintFn*) fn;
PgfCId name = *((PgfCId *) key);
PgfAbsFun *fun = *((PgfAbsFun **) value);
GuWriter *wtr = clo->wtr;
gu_puts(gu_seq_is_null(fun->defns) ? " data " : " fun ", wtr, err);
gu_string_write(name, wtr, err);
gu_puts(" : ", wtr, err);
pgf_print_type(fun->type, 0, wtr, err);
gu_printf(wtr, err, " ; -- %f\n", fun->ep.prob);
}
static void
pgf_print_abstract(PgfAbstr* abstr, GuWriter* wtr, GuExn* err)
{
gu_puts("abstract ", wtr, err);
gu_string_write(abstr->name, wtr, err);
gu_puts(" {\n", wtr, err);
PgfPrintFn clo1 = { { pgf_print_flag }, wtr };
gu_map_iter(abstr->aflags, &clo1.fn, err);
PgfPrintFn clo2 = { { pgf_print_cat }, wtr };
gu_map_iter(abstr->cats, &clo2.fn, err);
PgfPrintFn clo3 = { { pgf_print_absfun }, wtr };
gu_map_iter(abstr->funs, &clo3.fn, err);
gu_puts("}\n", wtr, err);
}
void
pgf_print_cat(GuMapItor* fn, const void* key, void* value,
GuExn* err)
{
PgfPrintFn* clo = (PgfPrintFn*) fn;
PgfCId name = *((PgfCId *) key);
PgfAbsCat *cat = *((PgfAbsCat **) value);
GuWriter *wtr = clo->wtr;
gu_puts(" cat ", wtr, err);
gu_string_write(name, wtr, err);
size_t n_hypos = gu_seq_length(cat->context);
for (size_t i = 0; i < n_hypos; i++) {
PgfHypo* hypo = gu_seq_get(cat->context, PgfHypo*, i);
gu_putc(' ', wtr, err);
pgf_print_hypo(hypo, 4, wtr, err);
}
gu_printf(wtr, err, " ; -- %f\n",cat->meta_prob);
}
static void
pgf_print_concrete(PgfCId cncname, PgfConcr* concr,
GuWriter* wtr, GuExn* err)
{
gu_puts("concrete ", wtr, err);
gu_string_write(cncname, wtr, err);
gu_puts(" {\n", wtr, err);
PgfPrintFn clo1 = { { pgf_print_flag }, wtr };
gu_map_iter(concr->cflags, &clo1.fn, err);
gu_puts(" productions\n", wtr, err);
PgfPrintFn clo2 = { { pgf_print_productions }, wtr };
gu_map_iter(concr->ccats, &clo2.fn, err);
gu_puts(" lindefs\n", wtr, err);
PgfPrintFn clo3 = { { pgf_print_lindefs }, wtr };
gu_map_iter(concr->ccats, &clo3.fn, err);
gu_puts(" lin\n", wtr, err);
size_t n_funs = gu_list_length(concr->cncfuns);
for (size_t i = 0; i < n_funs; i++) {
PgfCncFun* cncfun = gu_list_index(concr->cncfuns, i);
pgf_print_cncfun(cncfun, concr->sequences, wtr, err);
}
gu_puts(" sequences\n", wtr, err);
size_t n_seqs = gu_list_length(concr->sequences);
for (size_t i = 0; i < n_seqs; i++) {
PgfSequence seq = gu_list_index(concr->sequences, i);
pgf_print_sequence(i, seq, wtr, err);
}
gu_puts(" categories\n", wtr, err);
PgfPrintFn clo4 = { { pgf_print_cnccat }, wtr };
gu_map_iter(concr->cnccats, &clo4.fn, err);
gu_puts("}\n", wtr, err);
}
static void
pgf_print_cncfun(PgfCncFun *cncfun, PgfSequences *sequences,
GuWriter *wtr, GuExn *err)
{
gu_printf(wtr,err," F%d := (", cncfun->funid);
size_t n_seqs = gu_list_length(sequences);
for (size_t i = 0; i < cncfun->n_lins; i++) {
if (i > 0) gu_putc(',', wtr, err);
PgfSequence seq = cncfun->lins[i];
for (size_t seqid = 0; seqid < n_seqs; seqid++) {
if (gu_seq_data(gu_list_index(sequences, seqid)) == gu_seq_data(seq)) {
gu_printf(wtr,err,"S%d", seqid);
break;
}
}
}
gu_puts(") [", wtr, err);
gu_string_write(cncfun->name, wtr, err);
gu_puts("]\n", wtr, err);
}
int main(int argc, char* argv[]) {
// Set the character locale, so we can produce proper output.
setlocale(LC_CTYPE, "");
// Create the pool that is used to allocate everything
GuPool* pool = gu_new_pool();
int status = EXIT_SUCCESS;
if (argc < 5 || argc > 6) {
fprintf(stderr, "usage: %s pgf cat from-lang to-lang [probs-file]\n", argv[0]);
status = EXIT_FAILURE;
goto fail;
}
char* filename = argv[1];
GuString cat = gu_str_string(argv[2], pool);
GuString from_lang = gu_str_string(argv[3], pool);
GuString to_lang = gu_str_string(argv[4], pool);
// Create an exception frame that catches all errors.
GuExn* err = gu_new_exn(NULL, gu_kind(type), pool);
// Read the PGF grammar.
PgfPGF* pgf = pgf_read(filename, pool, err);
// If an error occured, it shows in the exception frame
if (!gu_ok(err)) {
fprintf(stderr, "Reading PGF failed\n");
status = EXIT_FAILURE;
goto fail;
}
if (argc == 6) {
char* meta_probs_filename = argv[5];
pgf_load_meta_child_probs(pgf, meta_probs_filename, pool, err);
if (!gu_ok(err)) {
fprintf(stderr, "Loading meta child probs failed\n");
status = EXIT_FAILURE;
goto fail;
}
}
// Look up the source and destination concrete categories
PgfConcr* from_concr = pgf_get_language(pgf, from_lang);
PgfConcr* to_concr = pgf_get_language(pgf, to_lang);
if (!from_concr || !to_concr) {
fprintf(stderr, "Unknown language\n");
status = EXIT_FAILURE;
goto fail_concr;
}
// Register a callback for the literal category Symbol
pgf_parser_add_literal(from_concr, gu_str_string("Symb", pool),
&pgf_nerc_literal_callback);
// Create an output stream for stdout
GuOut* out = gu_file_out(stdout, pool);
// Locale-encoding writers are currently unsupported
// GuWriter* wtr = gu_locale_writer(out, pool);
// Use a writer with hard-coded utf-8 encoding for now.
GuWriter* wtr = gu_new_utf8_writer(out, pool);
// We will keep the latest results in the 'ppool' and
// we will iterate over them by using 'result'.
GuPool* ppool = NULL;
GuEnum* result = NULL;
// The interactive translation loop.
// XXX: This currently reads stdin directly, so it doesn't support
// encodings properly. TODO: use a locale reader for input
while (true) {
fprintf(stdout, "> ");
fflush(stdout);
char buf[4096];
char* line = fgets(buf, sizeof(buf), stdin);
if (line == NULL) {
if (ferror(stdin)) {
fprintf(stderr, "Input error\n");
status = EXIT_FAILURE;
}
break;
} else if (strcmp(line, "") == 0) {
// End nicely on empty input
break;
} else if (strcmp(line, "\n") == 0) {
// Empty line -> show the next tree for the last sentence
if (result != NULL) {
clock_t start = clock();
PgfExprProb* ep = gu_next(result, PgfExprProb*, ppool);
clock_t end = clock();
double cpu_time_used = ((double) (end - start)) / CLOCKS_PER_SEC;
printf("%.2f sec\n", cpu_time_used);
// The enumerator will return a null variant at the
// end of the results.
if (ep == NULL) {
goto fail_parse;
}
print_result(ep, to_concr, wtr, err, ppool);
}
continue;
}
// We release the last results
if (ppool != NULL) {
gu_pool_free(ppool);
ppool = NULL;
result = NULL;
}
// We create a temporary pool for translating a single
// sentence, so our memory usage doesn't increase over time.
ppool = gu_new_pool();
GuReader *rdr =
gu_string_reader(gu_str_string(line, ppool), ppool);
PgfLexer *lexer =
pgf_new_lexer(rdr, ppool);
clock_t start = clock();
GuEnum* result =
pgf_parse(from_concr, cat, lexer, ppool);
if (result == NULL) {
PgfToken tok =
pgf_lexer_current_token(lexer);
if (gu_string_eq(tok, gu_empty_string))
gu_puts("Couldn't begin parsing", wtr, err);
else {
gu_puts("Unexpected token: \"", wtr, err);
gu_string_write(tok, wtr, err);
gu_puts("\"\n", wtr, err);
}
goto fail_parse;
}
PgfExprProb* ep = gu_next(result, PgfExprProb*, ppool);
clock_t end = clock();
double cpu_time_used = ((double) (end - start)) / CLOCKS_PER_SEC;
printf("%.2f sec\n", cpu_time_used);
// The enumerator will return null at the end of the results.
if (ep == NULL) {
goto fail_parse;
}
print_result(ep, to_concr, wtr, err, ppool);
continue;
fail_parse:
// Free all resources allocated during parsing and linearization
gu_pool_free(ppool);
ppool = NULL;
result = NULL;
}
int main(int argc, char* argv[]) {
// Set the character locale, so we can produce proper output.
setlocale(LC_CTYPE, "");
// Create the pool that is used to allocate everything
GuPool* pool = gu_new_pool();
int status = EXIT_SUCCESS;
if (argc < 5) {
fprintf(stderr, "usage: %s pgf cat from-lang to-lang\n", argv[0]);
status = EXIT_FAILURE;
goto fail;
}
GuString filename = argv[1];
GuString cat = argv[2];
GuString from_lang = argv[3];
GuString to_lang = argv[4];
// Create an exception frame that catches all errors.
GuExn* err = gu_new_exn(pool);
// Read the PGF grammar.
PgfPGF* pgf = pgf_read(filename, pool, err);
// If an error occured, it shows in the exception frame
if (!gu_ok(err)) {
fprintf(stderr, "Reading PGF failed\n");
status = EXIT_FAILURE;
goto fail;
}
// Look up the source and destination concrete categories
PgfConcr* from_concr = pgf_get_language(pgf, from_lang);
PgfConcr* to_concr = pgf_get_language(pgf, to_lang);
if (!from_concr || !to_concr) {
fprintf(stderr, "Unknown language\n");
status = EXIT_FAILURE;
goto fail_concr;
}
// Register a callback for the literal category Symbol
PgfCallbacksMap* callbacks =
pgf_new_callbacks_map(from_concr, pool);
pgf_callbacks_map_add_literal(from_concr, callbacks,
"PN", &pgf_nerc_literal_callback);
pgf_callbacks_map_add_literal(from_concr, callbacks,
"Symb", &pgf_unknown_literal_callback);
// Create an output stream for stdout
GuOut* out = gu_file_out(stdout, pool);
// We will keep the latest results in the 'ppool' and
// we will iterate over them by using 'result'.
GuPool* ppool = NULL;
GuEnum* result = NULL;
// The interactive translation loop.
// XXX: This currently reads stdin directly, so it doesn't support
// encodings properly. TODO: use a locale reader for input
while (true) {
fprintf(stdout, "> ");
fflush(stdout);
char buf[4096];
char* line = fgets(buf, sizeof(buf), stdin);
if (line == NULL) {
if (ferror(stdin)) {
fprintf(stderr, "Input error\n");
status = EXIT_FAILURE;
}
break;
} else if (strcmp(line, "") == 0) {
// End nicely on empty input
break;
} else if (strcmp(line, "\n") == 0) {
// Empty line -> show the next tree for the last sentence
if (result != NULL) {
clock_t start = clock();
PgfExprProb* ep = gu_next(result, PgfExprProb*, ppool);
clock_t end = clock();
double cpu_time_used = ((double) (end - start)) / CLOCKS_PER_SEC;
printf("%.2f sec\n", cpu_time_used);
// The enumerator will return a null variant at the
// end of the results.
if (ep == NULL) {
goto fail_parse;
}
print_result(ep, to_concr, out, err, ppool);
}
continue;
}
// We release the last results
if (ppool != NULL) {
gu_pool_free(ppool);
ppool = NULL;
result = NULL;
}
// We create a temporary pool for translating a single
// sentence, so our memory usage doesn't increase over time.
ppool = gu_new_pool();
clock_t start = clock();
GuExn* parse_err = gu_new_exn(ppool);
result =
pgf_parse_with_heuristics(from_concr, cat, line,
-1, callbacks,
parse_err, ppool, ppool);
if (!gu_ok(parse_err)) {
if (gu_exn_caught(parse_err, PgfExn)) {
GuString msg = gu_exn_caught_data(parse_err);
gu_string_write(msg, out, err);
gu_putc('\n', out, err);
} else if (gu_exn_caught(parse_err, PgfParseError)) {
gu_puts("Unexpected token: \"", out, err);
GuString tok = gu_exn_caught_data(parse_err);
gu_string_write(tok, out, err);
gu_puts("\"\n", out, err);
}
goto fail_parse;
}
PgfExprProb* ep = gu_next(result, PgfExprProb*, ppool);
clock_t end = clock();
double cpu_time_used = ((double) (end - start)) / CLOCKS_PER_SEC;
printf("%.2f sec\n", cpu_time_used);
// The enumerator will return null at the end of the results.
if (ep == NULL) {
goto fail_parse;
}
print_result(ep, to_concr, out, err, ppool);
continue;
fail_parse:
// Free all resources allocated during parsing and linearization
gu_pool_free(ppool);
ppool = NULL;
result = NULL;
}
