Solution, based on pdfextact.c seems like bruteforce, but it works:
- itarate through all pdf objects (
pdf_load_object
)
- determine if object is embedded file (
isembed
)
- if it is - access it's stream and save file (
saveembed
)
Embedded files stored at the end of file in most test cases, so, reverce iteration makes sence.
static int isembed(pdf_obj *obj) {
pdf_obj *type = pdf_dict_gets(obj, "Type");
return pdf_is_name(type) && !strcmp(pdf_to_name(type), "Filespec");
}
static void saveembed(pdf_obj *dict) {
char *filename;
pdf_obj *obj = pdf_dict_gets(dict, "F");
if (obj) filename = pdf_to_str_buf(obj);
obj = pdf_dict_gets(dict, "EF");
if (!obj) return;
pdf_obj *stream = pdf_dict_gets(obj, "F");
if (!stream) return;
FILE *f;
fz_buffer *buf;
int n, len;
unsigned char *data;
buf = pdf_load_stream(doc, pdf_to_num(stream), pdf_to_gen(stream));
printf("extracting embedded file %s\n", filename);
f = fopen(filename, "wb");
len = fz_buffer_storage(ctx, buf, &data);
n = fwrite(data, 1, len, f);
fclose(f);
fz_drop_buffer(ctx, buf);
}