static int
csv_accept(strm_stream* strm, strm_value data)
{
strm_array ary;
strm_string line = strm_value_str(data);
strm_value *bp;
const char *fbeg;
const char *ptr;
const char *pend;
int fieldcnt;
int in_quote = 0, all_str = 1;
int i = 0;
enum csv_type ftype;
enum csv_type* types;
struct csv_data *cd = strm->data;
if (cd->prev) {
strm_int len = strm_str_len(cd->prev)+strm_str_len(line)+1;
char* tmp = malloc(len);
memcpy(tmp, strm_str_ptr(cd->prev), strm_str_len(cd->prev));
*(tmp+strm_str_len(cd->prev)) = '\n';
memcpy(tmp+strm_str_len(cd->prev)+1, strm_str_ptr(line), strm_str_len(line));
line = strm_str_new(tmp, len);
free(tmp);
cd->prev = strm_str_null;
}
fieldcnt = count_fields(line);
if (fieldcnt == -1) {
cd->prev = line;
return STRM_NG;
}
if (cd->n > 0 && fieldcnt != cd->n)
return STRM_NG;
ptr = strm_str_ptr(line);
pend = ptr + strm_str_len(line);
ary = strm_ary_new(NULL, fieldcnt);
if (!ary) return STRM_NG;
bp = (strm_value*)strm_ary_ptr(ary);
types = cd->types;
ftype = types ? types[0] : TYPE_UNSPC;
for (fbeg=ptr; ptr<pend; ptr++) {
if (in_quote) {
if (*ptr == '\"') {
if (ptr[1] == '\"') {
ptr++;
ftype = TYPE_ESC;
continue;
}
in_quote = 0;
}
continue;
}
switch(*ptr) {
case '\"':
in_quote = 1;
if (ptr == fbeg) {
ftype = TYPE_STR;
fbeg = ptr+1;
}
else {
ftype = TYPE_ESC;
}
continue;
case ',':
*bp = csv_value(fbeg, ptr-fbeg, ftype);
if (!strm_string_p(*bp)) all_str = 0;
bp++;
fbeg = ptr+1;
i++;
ftype = types ? types[i] : TYPE_UNSPC;
break;
default:
continue;
}
}
/* trim newline at the end */
if (ptr[-1] == '\n') {
ptr--;
}
/* trim carriage return at the end */
if (ptr[-1] == '\r') {
ptr--;
}
*bp = csv_value(fbeg, ptr-fbeg, ftype);
if (!strm_string_p(*bp)) all_str = 0;
/* check headers */
if (!cd->headers && !cd->types) {
if (all_str) {
cd->headers = ary;
ary = strm_ary_null;
}
cd->n = fieldcnt;
}
if (ary) {
/* set headers if any */
if (cd->headers)
strm_ary_headers(ary) = cd->headers;
if (!cd->types) {
/* first data line (after optinal header line) */
if (cd->headers) {
if (all_str) { /* data line is all string; emit header line */
strm_emit(strm, strm_ary_value(cd->headers), NULL);
cd->headers = strm_ary_null;
}
else { /* intern header strings */
strm_array h = cd->headers;
strm_value *p = strm_ary_ptr(h);
int i;
for (i=0; i<strm_ary_len(h); i++) {
strm_string str = strm_value_str(p[i]);
p[i] = strm_str_value(strm_str_intern_str(str));
}
}
}
/* initialize types (determined by first data line) */
cd->types = malloc(sizeof(enum csv_type)*fieldcnt);
if (!cd->types) return STRM_NG;
for (i=0; i<fieldcnt; i++) {
cd->types[i] = csv_type(strm_ary_ptr(ary)[i]);
}
}
else {
/* type check */
for (i=0; i<fieldcnt; i++) {
if (cd->types[i] != csv_type(strm_ary_ptr(ary)[i])) {
/* type mismatch (error); skip this line */
strm_raise(strm, "csv type mismatch");
return STRM_NG;
}
}
}
strm_emit(strm, strm_str_value(ary), NULL);
}
return STRM_OK;
}
static void
csv_accept(strm_task* task, strm_value data)
{
strm_array *ary;
strm_string *line = strm_value_str(data);
strm_value *bp;
char *tmp, *tptr;
const char *ptr;
const char *pend;
int fieldcnt, len;
int in_quote = 0, quoted = 0, all_str = 1;;
struct csv_data *cd = task->data;
if (cd->prev) {
strm_string *str = strm_str_new(NULL, cd->prev->len+line->len+1);
tmp = (char*)str->ptr;
memcpy(tmp, cd->prev->ptr, cd->prev->len);
*(tmp+cd->prev->len) = '\n';
memcpy(tmp+cd->prev->len+1, line->ptr, line->len);
line = str;
cd->prev = NULL;
}
fieldcnt = count_fields(line);
if (fieldcnt == -1) {
cd->prev = line;
return;
}
if (cd->n > 0 && fieldcnt != cd->n)
return;
ptr = line->ptr;
pend = ptr + line->len;
ary = strm_ary_new(NULL, fieldcnt);
if (!ary) return;
bp = (strm_value*)ary->ptr;
len = line->len;
tmp = malloc(len+1);
if (!tmp) return;
*tmp='\0';
ptr=line->ptr;
tptr=tmp;
for (;ptr<pend; ptr++) {
if (in_quote) {
if (*ptr == '\"') {
if (ptr[1] == '\"') {
*tptr++ = '\"';
ptr++;
continue;
}
in_quote = 0;
}
else
*tptr++ = *ptr;
continue;
}
switch(*ptr) {
case '\"':
in_quote = 1;
quoted = 1;
continue;
case ',':
if (quoted) {
*bp = strm_str_value(tmp, tptr-tmp);
}
else {
*bp = csv_value(tmp, tptr-tmp);
}
if (!strm_str_p(*bp)) all_str = 0;
bp++;
tptr = tmp;
quoted = 0;
break;
default:
*tptr++ = *ptr;
continue;
}
}
/* trim newline at the end */
if (tptr > tmp && tptr[-1] == '\n') {
tptr--;
}
/* trim carriage return at the end */
if (tptr > tmp && tptr[-1] == '\r') {
tptr--;
}
*bp = csv_value(tmp, tptr-tmp);
if (!strm_str_p(*bp)) all_str = 0;
free(tmp);
/* check headers */
if (!cd->headers && !cd->types) {
if (all_str) {
cd->headers = ary;
ary = NULL;
}
cd->n = fieldcnt;
}
if (ary) {
int i;
/* set headers if any */
if (cd->headers)
ary->headers = cd->headers;
if (!cd->types) {
/* first data line (after optinal header line) */
if (cd->headers) {
if (all_str) { /* data line is all string; emit header line */
strm_emit(task, strm_ptr_value(cd->headers), NULL);
cd->headers = NULL;
}
else { /* intern header strings */
strm_array *h = cd->headers;
strm_value *p = (strm_value*)h->ptr;
int i;
for (i=0; i<h->len; i++) {
strm_string *str = strm_value_str(p[i]);
p[i] = strm_ptr_value(strm_str_intern_str(str));
}
}
}
/* initialize types (determined by first data line) */
cd->types = malloc(sizeof(enum csv_type)*fieldcnt);
if (!cd->types) return;
for (i=0; i<fieldcnt; i++) {
cd->types[i] = csv_type(ary->ptr[i]);
}
}
else {
/* type check */
for (i=0; i<fieldcnt; i++) {
if (cd->types[i] != csv_type(ary->ptr[i])) {
if (cd->types[i] == STRING_TYPE) {
/* convert value to string */
((strm_value*)ary->ptr)[i] = strm_ptr_value(strm_to_str(ary->ptr[i]));
}
else {
/* type mismatch (error); skip this line */
return;
}
}
}
}
strm_emit(task, strm_ptr_value(ary), NULL);
}
}