rag/app/naive.py

#  Licensed under the Apache License, Version 2.0 (the "License");
#  you may not use this file except in compliance with the License.
#  You may obtain a copy of the License at
#
#      http://www.apache.org/licenses/LICENSE-2.0
#
#  Unless required by applicable law or agreed to in writing, software
#  distributed under the License is distributed on an "AS IS" BASIS,
#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
#  See the License for the specific language governing permissions and
#  limitations under the License.
#
from tika import parser
from io import BytesIO
from docx import Document
from timeit import default_timer as timer
import re
from deepdoc.parser.pdf_parser import PlainParser
from rag.nlp import rag_tokenizer, naive_merge, tokenize_table, tokenize_chunks, find_codec, concat_img, naive_merge_docx, tokenize_chunks_docx
from deepdoc.parser import PdfParser, ExcelParser, DocxParser, HtmlParser, JsonParser, MarkdownParser, TxtParser
from rag.settings import cron_logger
from rag.utils import num_tokens_from_string
from PIL import Image
from functools import reduce
from markdown import markdown
from docx.image.exceptions import UnrecognizedImageError

class Docx(DocxParser):
    def __init__(self):
        pass

    def get_picture(self, document, paragraph):
        img = paragraph._element.xpath('.//pic:pic')
        if not img:
            return None
        img = img[0]
        embed = img.xpath('.//a:blip/@r:embed')[0]
        related_part = document.part.related_parts[embed]
        try:
            image_blob = related_part.image.blob
        except UnrecognizedImageError:
            print("Unrecognized image format. Skipping image.")
            return None
        try:
            image = Image.open(BytesIO(image_blob)).convert('RGB')
            return image
        except Exception as e:
            return None

    def __clean(self, line):
        line = re.sub(r"\u3000", " ", line).strip()
        return line

    def __call__(self, filename, binary=None, from_page=0, to_page=100000):
        self.doc = Document(
            filename) if not binary else Document(BytesIO(binary))
        pn = 0
        lines = []
        last_image = None
        for p in self.doc.paragraphs:
            if pn > to_page:
                break
            if from_page <= pn < to_page:
                if p.text.strip():
                    if p.style and p.style.name == 'Caption':
                        former_image = None
                        if lines and lines[-1][1] and lines[-1][2] != 'Caption':
                            former_image = lines[-1][1].pop()
                        elif last_image:
                            former_image = last_image
                            last_image = None
                        lines.append((self.__clean(p.text), [former_image], p.style.name))
                    else:
                        current_image = self.get_picture(self.doc, p)
                        image_list = [current_image]
                        if last_image:
                            image_list.insert(0, last_image)
                            last_image = None
                        lines.append((self.__clean(p.text), image_list, p.style.name))
                else:
                    if current_image := self.get_picture(self.doc, p):
                        if lines:
                            lines[-1][1].append(current_image)
                        else:
                            last_image = current_image
            for run in p.runs:
                if 'lastRenderedPageBreak' in run._element.xml:
                    pn += 1
                    continue
                if 'w:br' in run._element.xml and 'type="page"' in run._element.xml:
                    pn += 1
        new_line = [(line[0], reduce(concat_img, line[1]) if line[1] else None) for line in lines]

        tbls = []
        for tb in self.doc.tables:
            html= "<table>"
            for r in tb.rows:
                html += "<tr>"
                i = 0
                while i < len(r.cells):
                    span = 1
                    c = r.cells[i]
                    for j in range(i+1, len(r.cells)):
                        if c.text == r.cells[j].text:
                            span += 1
                            i = j
                    i += 1
                    html += f"<td>{c.text}</td>" if span == 1 else f"<td colspan='{span}'>{c.text}</td>"
                html += "</tr>"
            html += "</table>"
            tbls.append(((None, html), ""))
        return new_line, tbls


class Pdf(PdfParser):
    def __call__(self, filename, binary=None, from_page=0,
                 to_page=100000, zoomin=3, callback=None):
        start = timer()
        callback(msg="OCR is running...")
        self.__images__(
            filename if not binary else binary,
            zoomin,
            from_page,
            to_page,
            callback
        )
        callback(msg="OCR finished")
        cron_logger.info("OCR({}~{}): {}".format(from_page, to_page, timer() - start))

        start = timer()
        self._layouts_rec(zoomin)
        callback(0.63, "Layout analysis finished.")
        self._table_transformer_job(zoomin)
        callback(0.65, "Table analysis finished.")
        self._text_merge()
        callback(0.67, "Text merging finished")
        tbls = self._extract_table_figure(True, zoomin, True, True)
        #self._naive_vertical_merge()
        self._concat_downward()
        #self._filter_forpages()

        cron_logger.info("layouts: {}".format(timer() - start))
        return [(b["text"], self._line_tag(b, zoomin))
                for b in self.boxes], tbls


class Markdown(MarkdownParser):
    def __call__(self, filename, binary=None):
        txt = ""
        tbls = []
        if binary:
            encoding = find_codec(binary)
            txt = binary.decode(encoding, errors="ignore")
        else:
            with open(filename, "r") as f:
                txt = f.read()
        remainder, tables = self.extract_tables_and_remainder(f'{txt}\n')
        sections = []
        tbls = []
        for sec in remainder.split("\n"):
            if num_tokens_from_string(sec) > 10 * self.chunk_token_num:
                sections.append((sec[:int(len(sec)/2)], ""))
                sections.append((sec[int(len(sec)/2):], ""))
            else:
                sections.append((sec, ""))
        print(tables)
        for table in tables:
            tbls.append(((None, markdown(table, extensions=['markdown.extensions.tables'])), ""))
        return sections, tbls



def chunk(filename, binary=None, from_page=0, to_page=100000,
          lang="Chinese", callback=None, **kwargs):
    """
        Supported file formats are docx, pdf, excel, txt.
        This method apply the naive ways to chunk files.
        Successive text will be sliced into pieces using 'delimiter'.
        Next, these successive pieces are merge into chunks whose token number is no more than 'Max token number'.
    """

    eng = lang.lower() == "english"  # is_english(cks)
    parser_config = kwargs.get(
        "parser_config", {
            "chunk_token_num": 128, "delimiter": "\n!?。；！？", "layout_recognize": True})
    doc = {
        "docnm_kwd": filename,
        "title_tks": rag_tokenizer.tokenize(re.sub(r"\.[a-zA-Z]+$", "", filename))
    }
    doc["title_sm_tks"] = rag_tokenizer.fine_grained_tokenize(doc["title_tks"])
    res = []
    pdf_parser = None
    sections = []
    if re.search(r"\.docx$", filename, re.IGNORECASE):
        callback(0.1, "Start to parse.")
        sections, tbls = Docx()(filename, binary)
        res = tokenize_table(tbls, doc, eng)    # just for table

        callback(0.8, "Finish parsing.")
        st = timer()

        chunks, images = naive_merge_docx(
            sections, int(parser_config.get(
                "chunk_token_num", 128)), parser_config.get(
                "delimiter", "\n!?。；！？"))

        if kwargs.get("section_only", False):
            return chunks

        res.extend(tokenize_chunks_docx(chunks, doc, eng, images))
        cron_logger.info("naive_merge({}): {}".format(filename, timer() - st))
        return res

    elif re.search(r"\.pdf$", filename, re.IGNORECASE):
        pdf_parser = Pdf(
        ) if parser_config.get("layout_recognize", True) else PlainParser()
        sections, tbls = pdf_parser(filename if not binary else binary,
                                    from_page=from_page, to_page=to_page, callback=callback)
        res = tokenize_table(tbls, doc, eng)

    elif re.search(r"\.xlsx?$", filename, re.IGNORECASE):
        callback(0.1, "Start to parse.")
        excel_parser = ExcelParser()
        sections = [(l, "") for l in excel_parser.html(binary) if l]

    elif re.search(r"\.(txt|py|js|java|c|cpp|h|php|go|ts|sh|cs|kt|sql)$", filename, re.IGNORECASE):
        callback(0.1, "Start to parse.")
        sections = TxtParser()(filename,binary,
                               parser_config.get("chunk_token_num", 128),
                               parser_config.get("delimiter", "\n!?;。；！？"))
        callback(0.8, "Finish parsing.")
    
    elif re.search(r"\.(md|markdown)$", filename, re.IGNORECASE):
        callback(0.1, "Start to parse.")
        sections, tbls = Markdown(int(parser_config.get("chunk_token_num", 128)))(filename, binary)
        res = tokenize_table(tbls, doc, eng)
        callback(0.8, "Finish parsing.")

    elif re.search(r"\.(htm|html)$", filename, re.IGNORECASE):
        callback(0.1, "Start to parse.")
        sections = HtmlParser()(filename, binary)
        sections = [(l, "") for l in sections if l]
        callback(0.8, "Finish parsing.")

    elif re.search(r"\.json$", filename, re.IGNORECASE):
        callback(0.1, "Start to parse.")
        sections = JsonParser(int(parser_config.get("chunk_token_num", 128)))(binary)
        sections = [(l, "") for l in sections if l]
        callback(0.8, "Finish parsing.")

    elif re.search(r"\.doc$", filename, re.IGNORECASE):
        callback(0.1, "Start to parse.")
        binary = BytesIO(binary)
        doc_parsed = parser.from_buffer(binary)
        sections = doc_parsed['content'].split('\n')
        sections = [(l, "") for l in sections if l]
        callback(0.8, "Finish parsing.")

    else:
        raise NotImplementedError(
            "file type not supported yet(pdf, xlsx, doc, docx, txt supported)")

    st = timer()
    chunks = naive_merge(
        sections, int(parser_config.get(
            "chunk_token_num", 128)), parser_config.get(
            "delimiter", "\n!?。；！？"))
    if kwargs.get("section_only", False):
        return chunks

    res.extend(tokenize_chunks(chunks, doc, eng, pdf_parser))
    cron_logger.info("naive_merge({}): {}".format(filename, timer() - st))
    return res


if __name__ == "__main__":
    import sys

    def dummy(prog=None, msg=""):
        pass

    chunk(sys.argv[1], from_page=0, to_page=10, callback=dummy)