Files
sist2/src/parsing/doc.c
2020-02-07 10:08:01 -05:00

130 lines
3.8 KiB
C

#include "doc.h"
#include "src/ctx.h"
int dump_text(mceTextReader_t *reader, dyn_buffer_t *buf) {
mce_skip_attributes(reader);
xmlErrorPtr err = xmlGetLastError();
if (err != NULL) {
if (err->level == XML_ERR_FATAL) {
LOG_ERRORF("doc.c", "Got fatal XML error while parsing document: %s", err->message)
return -1;
} else {
LOG_ERRORF("doc.c", "Got recoverable XML error while parsing document: %s", err->message)
}
}
mce_start_children(reader) {
mce_start_element(reader, NULL, _X("t")) {
mce_skip_attributes(reader);
mce_start_children(reader) {
mce_start_text(reader) {
char *str = (char *) xmlTextReaderConstValue(reader->reader);
dyn_buffer_append_string(buf, str);
dyn_buffer_write_char(buf, ' ');
} mce_end_text(reader);
} mce_end_children(reader);
} mce_end_element(reader);
mce_start_element(reader, NULL, NULL) {
int ret = dump_text(reader, buf);
if (ret != 0) {
return ret;
}
} mce_end_element(reader);
} mce_end_children(reader)
return 0;
}
__always_inline
int should_read_part(opcPart part) {
char *part_name = (char *) part;
if (part == NULL) {
return FALSE;
}
if ( // Word
strcmp(part_name, "word/document.xml") == 0
|| strncmp(part_name, "word/footer", sizeof("word/footer") - 1) == 0
|| strncmp(part_name, "word/header", sizeof("word/header") - 1) == 0
// PowerPoint
|| strncmp(part_name, "ppt/slides/slide", sizeof("ppt/slides/slide") - 1) == 0
|| strncmp(part_name, "ppt/notesSlides/notesSlide", sizeof("ppt/notesSlides/notesSlide") - 1) == 0
// Excel
|| strncmp(part_name, "xl/worksheets/sheet", sizeof("xl/worksheets/sheet") - 1) == 0
|| strcmp(part_name, "xl/sharedStrings.xml") == 0
|| strcmp(part_name, "xl/workbook.xml") == 0
) {
return TRUE;
}
return FALSE;
}
__always_inline
int read_part(opcContainer *c, dyn_buffer_t *buf, opcPart part, document_t *doc) {
mceTextReader_t reader;
int ret = opcXmlReaderOpen(c, &reader, part, NULL, "UTF-8", XML_PARSE_NOWARNING | XML_PARSE_NOERROR | XML_PARSE_NONET);
if (ret != OPC_ERROR_NONE) {
LOG_ERRORF(doc->filepath, "(doc.c) opcXmlReaderOpen() returned error code %d", ret);
return -1;
}
mce_start_document(&reader) {
mce_start_element(&reader, NULL, NULL) {
ret = dump_text(&reader, buf);
if (ret != 0) {
mceTextReaderCleanup(&reader);
return -1;
}
} mce_end_element(&reader);
} mce_end_document(&reader);
mceTextReaderCleanup(&reader);
return 0;
}
void parse_doc(void *mem, size_t mem_len, document_t *doc) {
if (mem == NULL) {
return;
}
opcContainer *c = opcContainerOpenMem(mem, mem_len, OPC_OPEN_READ_ONLY, NULL);
if (c == NULL) {
LOG_ERROR(doc->filepath, "(doc.c) Couldn't open document with opcContainerOpenMem()");
return;
}
dyn_buffer_t buf = dyn_buffer_create();
opcPart part = opcPartGetFirst(c);
do {
if (should_read_part(part)) {
int ret = read_part(c, &buf, part, doc);
if (ret != 0) {
break;
}
}
} while ((part = opcPartGetNext(c, part)));
opcContainerClose(c, OPC_CLOSE_NOW);
if (buf.cur > 0) {
dyn_buffer_write_char(&buf, '\0');
meta_line_t *meta = malloc(sizeof(meta_line_t) + buf.cur);
meta->key = MetaContent;
strcpy(meta->strval, buf.buf);
APPEND_META(doc, meta)
}
dyn_buffer_destroy(&buf);
}