static char* read_line_pfr_psb(peek_file_reader_t* pfr, string_builder_t* psb, parse_trie_t* ptrie) {
int rc, stridx, matchlen;
while (TRUE) {
pfr_buffer_by(pfr, ptrie->maxlen);
rc = parse_trie_ring_match(ptrie, pfr->peekbuf, pfr->sob, pfr->npeeked, pfr->peekbuflenmask,
&stridx, &matchlen);
if (rc) {
pfr_advance_by(pfr, matchlen);
switch(stridx) {
case IRS_STRIDX:
return sb_finish(psb);
break;
case IRSEOF_STRIDX:
return sb_finish(psb);
break;
case EOF_STRIDX:
return NULL;
break;
}
} else {
sb_append_char(psb, pfr_read_char(pfr));
}
}
}
static int lrec_reader_stdio_csv_get_fields(lrec_reader_stdio_csv_state_t* pstate, rslls_t* pfields,
context_t* pctx, int is_header)
{
int rc, stridx, matchlen, record_done, field_done;
peek_file_reader_t* pfr = pstate->pfr;
string_builder_t* psb = pstate->psb;
char* field = NULL;
int field_length = 0;
if (pfr_peek_char(pfr) == (char)EOF) // char defaults to unsigned on some platforms
return FALSE;
// Strip the UTF-8 BOM, if any. This is MUCH simpler for mmap, and for stdio on files. For mmap
// we can test the first 3 bytes, then skip past them or not. For stdio on files we can fread
// the first 3 bytes, then rewind the fp if they're not the UTF-8 BOM. But for stdio on stdin
// (which is the primary reason we support stdio in Miller), we cannot rewind: stdin is not
// rewindable.
if (is_header) {
pfr_buffer_by(pfr, UTF8_BOM_LENGTH);
int rc = parse_trie_ring_match(pstate->putf8_bom_parse_trie,
pfr->peekbuf, pfr->sob, pfr->npeeked, pfr->peekbuflenmask,
&stridx, &matchlen);
#ifdef DEBUG_PARSER
printf("RC=%d stridx=0x%04x matchlen=%d\n", rc, stridx, matchlen);
#endif
if (rc == TRUE && stridx == UTF8_BOM_STRIDX) {
pfr_advance_by(pfr, matchlen);
}
}
// Loop over fields in record
record_done = FALSE;
while (!record_done) {
// Assumption is dquote is "\""
if (pfr_peek_char(pfr) != pstate->dquote[0]) { // NOT DOUBLE-QUOTED
// Loop over characters in field
field_done = FALSE;
while (!field_done) {
pfr_buffer_by(pfr, pstate->pno_dquote_parse_trie->maxlen);
rc = parse_trie_ring_match(pstate->pno_dquote_parse_trie,
pfr->peekbuf, pfr->sob, pfr->npeeked, pfr->peekbuflenmask,
&stridx, &matchlen);
#ifdef DEBUG_PARSER
pfr_print(pfr);
#endif
if (rc) {
#ifdef DEBUG_PARSER
printf("RC=%d stridx=0x%04x matchlen=%d\n", rc, stridx, matchlen);
#endif
switch(stridx) {
case EOF_STRIDX: // end of record
rslls_append(pfields, sb_finish(psb), FREE_ENTRY_VALUE, 0);
field_done = TRUE;
record_done = TRUE;
break;
case IFS_EOF_STRIDX:
fprintf(stderr, "%s: syntax error: record-ending field separator at line %lld.\n",
MLR_GLOBALS.bargv0, pstate->ilno);
exit(1);
break;
case IFS_STRIDX: // end of field
rslls_append(pfields, sb_finish(psb), FREE_ENTRY_VALUE, 0);
field_done = TRUE;
break;
case IRS_STRIDX: // end of record
field = sb_finish_with_length(psb, &field_length);
// The line-ending '\n' won't be included in the field buffer.
if (pstate->do_auto_line_term) {
if (field_length > 0 && field[field_length-1] == '\r') {
field[field_length-1] = 0;
context_set_autodetected_crlf(pctx);
} else {
context_set_autodetected_lf(pctx);
}
}
rslls_append(pfields, field, FREE_ENTRY_VALUE, 0);
field_done = TRUE;
record_done = TRUE;
break;
case DQUOTE_STRIDX: // CSV syntax error: fields containing quotes must be fully wrapped in quotes
fprintf(stderr, "%s: syntax error: unwrapped double quote at line %lld.\n",
MLR_GLOBALS.bargv0, pstate->ilno);
exit(1);
break;
default:
fprintf(stderr, "%s: internal coding error: unexpected token %d at line %lld.\n",
MLR_GLOBALS.bargv0, stridx, pstate->ilno);
exit(1);
break;
}
pfr_advance_by(pfr, matchlen);
} else {
#ifdef DEBUG_PARSER
char c = pfr_read_char(pfr);
printf("CHAR=%c [%02x]\n", isprint((unsigned char)c) ? c : ' ', (unsigned)c);
sb_append_char(psb, c);
#else
sb_append_char(psb, pfr_read_char(pfr));
#endif
}
}
} else { // DOUBLE-QUOTED
pfr_advance_by(pfr, pstate->dquotelen);
// loop over characters in field
field_done = FALSE;
char* field = NULL;
int field_length = 0;
while (!field_done) {
pfr_buffer_by(pfr, pstate->pdquote_parse_trie->maxlen);
rc = parse_trie_ring_match(pstate->pdquote_parse_trie,
pfr->peekbuf, pfr->sob, pfr->npeeked, pfr->peekbuflenmask,
&stridx, &matchlen);
if (rc) {
switch(stridx) {
case EOF_STRIDX: // end of record
fprintf(stderr, "%s: unmatched double quote at line %lld.\n",
MLR_GLOBALS.bargv0, pstate->ilno);
exit(1);
break;
case DQUOTE_EOF_STRIDX: // end of record
rslls_append(pfields, sb_finish(psb), FREE_ENTRY_VALUE, FIELD_QUOTED_ON_INPUT);
field_done = TRUE;
record_done = TRUE;
break;
case DQUOTE_IFS_STRIDX: // end of field
rslls_append(pfields, sb_finish(psb), FREE_ENTRY_VALUE, FIELD_QUOTED_ON_INPUT);
field_done = TRUE;
break;
case DQUOTE_IRS_STRIDX: // end of record
case DQUOTE_IRS2_STRIDX: // end of record
field = sb_finish_with_length(psb, &field_length);
// The line-ending '\n' won't be included in the field buffer.
if (pstate->do_auto_line_term) {
if (field_length > 0 && field[field_length-1] == '\r') {
field[field_length-1] = 0;
context_set_autodetected_crlf(pctx);
} else {
context_set_autodetected_lf(pctx);
}
}
rslls_append(pfields, field, FREE_ENTRY_VALUE, FIELD_QUOTED_ON_INPUT);
field_done = TRUE;
record_done = TRUE;
break;
case DQUOTE_DQUOTE_STRIDX: // RFC-4180 CSV: "" inside a dquoted field is an escape for "
sb_append_char(psb, pstate->dquote[0]);
break;
default:
fprintf(stderr, "%s: internal coding error: unexpected token %d at line %lld.\n",
MLR_GLOBALS.bargv0, stridx, pstate->ilno);
exit(1);
break;
}
pfr_advance_by(pfr, matchlen);
} else {
sb_append_char(psb, pfr_read_char(pfr));
}
}
}
}
return TRUE;
}