solve task execution issues (#90)

api/apps/document_app.py

     try:
         for id in req["doc_ids"]:
             info = {"run": str(req["run"]), "progress": 0}
-            if str(req["run"]) == TaskStatus.RUNNING.value:info["progress_msg"] = ""
+            if str(req["run"]) == TaskStatus.RUNNING.value:
+                info["progress_msg"] = ""
+                info["chunk_num"] = 0
+                info["token_num"] = 0
             DocumentService.update_by_id(id, info)
-            if str(req["run"]) == TaskStatus.CANCEL.value:
-                tenant_id = DocumentService.get_tenant_id(id)
-                if not tenant_id:
-                    return get_data_error_result(retmsg="Tenant not found!")
-                ELASTICSEARCH.deleteByQuery(Q("match", doc_id=id), idxnm=search.index_name(tenant_id))
+            #if str(req["run"]) == TaskStatus.CANCEL.value:
+            tenant_id = DocumentService.get_tenant_id(id)
+            if not tenant_id:
+                return get_data_error_result(retmsg="Tenant not found!")
+            ELASTICSEARCH.deleteByQuery(Q("match", doc_id=id), idxnm=search.index_name(tenant_id))
 
         return get_json_result(data=True)
     except Exception as e:
         if doc.type == FileType.VISUAL or re.search(r"\.(ppt|pptx|pages)$", doc.name):
             return get_data_error_result(retmsg="Not supported yet!")
 
-        e = DocumentService.update_by_id(doc.id, {"parser_id": req["parser_id"], "progress":0, "progress_msg": ""})
+        e = DocumentService.update_by_id(doc.id, {"parser_id": req["parser_id"], "progress":0, "progress_msg": "", "run": "0"})
         if not e:
             return get_data_error_result(retmsg="Document not found!")
         if doc.token_num>0:
             e = DocumentService.increment_chunk_num(doc.id, doc.kb_id, doc.token_num*-1, doc.chunk_num*-1, doc.process_duation*-1)
             if not e:
                 return get_data_error_result(retmsg="Document not found!")
+            tenant_id = DocumentService.get_tenant_id(req["doc_id"])
+            if not tenant_id:
+                return get_data_error_result(retmsg="Tenant not found!")
+            ELASTICSEARCH.deleteByQuery(Q("match", doc_id=doc.id), idxnm=search.index_name(tenant_id))
 
         return get_json_result(data=True)
     except Exception as e:

api/db/services/task_service.py

         try:
             task = cls.model.get_by_id(id)
             _, doc = DocumentService.get_by_id(task.doc_id)
-            return doc.run == TaskStatus.CANCEL.value
+            return doc.run == TaskStatus.CANCEL.value or doc.progress < 0
         except Exception as e:
             pass
         return True

api/settings.py

 
 DATABASE = decrypt_database_config(name="mysql")
 
-# Logger
-LoggerFactory.set_directory(os.path.join(get_project_base_directory(), "logs", "api"))
-# {CRITICAL: 50, FATAL:50, ERROR:40, WARNING:30, WARN:30, INFO:20, DEBUG:10, NOTSET:0}
-LoggerFactory.LEVEL = 10
-
-stat_logger = getLogger("stat")
-access_logger = getLogger("access")
-database_logger = getLogger("database")
-
 # Switch
 # upload
 UPLOAD_DATA_FROM_CLIENT = True
 
 retrievaler = search.Dealer(ELASTICSEARCH)
 
+# Logger
+LoggerFactory.set_directory(os.path.join(get_project_base_directory(), "logs", "api"))
+# {CRITICAL: 50, FATAL:50, ERROR:40, WARNING:30, WARN:30, INFO:20, DEBUG:10, NOTSET:0}
+LoggerFactory.LEVEL = 10
+
+stat_logger = getLogger("stat")
+access_logger = getLogger("access")
+database_logger = getLogger("database")
+
 class CustomEnum(Enum):
     @classmethod
     def valid(cls, value):

deepdoc/parser/pdf_parser.py

 import re
 import pdfplumber
 import logging
-from PIL import Image
+from PIL import Image, ImageDraw
 import numpy as np
 
 from api.db import ParserType
 
     def crop(self, text, ZM=3):
         imgs = []
+        poss = []
         for tag in re.findall(r"@@[0-9-]+\t[0-9.\t]+##", text):
             pn, left, right, top, bottom = tag.strip(
                 "#").strip("@").split("\t")
             left, right, top, bottom = float(left), float(
                 right), float(top), float(bottom)
+            poss.append(([int(p) - 1 for p in pn.split("-")], left, right, top, bottom))
+        if not poss: return
+
+        max_width = np.max([right-left for (_, left, right, _, _) in poss])
+        GAP = 6
+        pos = poss[0]
+        poss.insert(0, ([pos[0][0]], pos[1], pos[2], max(0, pos[3]-120), max(pos[3]-GAP, 0)))
+        pos = poss[-1]
+        poss.append(([pos[0][-1]], pos[1], pos[2], min(self.page_images[pos[0][-1]].size[1]/ZM, pos[4]+GAP), min(self.page_images[pos[0][-1]].size[1]/ZM, pos[4]+120)))
+
+        for ii, (pns, left, right, top, bottom) in enumerate(poss):
+            right = left + max_width
             bottom *= ZM
-            pns = [int(p) - 1 for p in pn.split("-")]
             for pn in pns[1:]:
                 bottom += self.page_images[pn - 1].size[1]
             imgs.append(
 
         if not imgs:
             return
-        GAP = 2
         height = 0
         for img in imgs:
             height += img.size[1] + GAP
         height = int(height)
+        width = int(np.max([i.size[0] for i in imgs]))
         pic = Image.new("RGB",
-                        (int(np.max([i.size[0] for i in imgs])), height),
+                        (width, height),
                         (245, 245, 245))
         height = 0
-        for img in imgs:
+        for ii, img in enumerate(imgs):
+            if ii == 0 or ii + 1 == len(imgs):
+                img = img.convert('RGBA')
+                overlay = Image.new('RGBA', img.size, (0, 0, 0, 0))
+                overlay.putalpha(128)
+                img = Image.alpha_composite(img, overlay).convert("RGB")
             pic.paste(img, (0, int(height)))
             height += img.size[1] + GAP
         return pic

deepdoc/vision/layout_recognizer.py

              "Equation",
         ]
     def __init__(self, domain):
-        super().__init__(self.labels, domain) #, os.path.join(get_project_base_directory(), "rag/res/deepdoc/"))
+        super().__init__(self.labels, domain, os.path.join(get_project_base_directory(), "rag/res/deepdoc/"))
 
     def __call__(self, image_list, ocr_res, scale_factor=3, thr=0.2, batch_size=16):
         def __is_garbage(b):

deepdoc/vision/table_structure_recognizer.py

     ]
 
     def __init__(self):
-        super().__init__(self.labels, "tsr")#,os.path.join(get_project_base_directory(), "rag/res/deepdoc/"))
+        super().__init__(self.labels, "tsr",os.path.join(get_project_base_directory(), "rag/res/deepdoc/"))
 
     def __call__(self, images, thr=0.2):
         tbls = super().__call__(images, thr)

rag/app/book.py

 import copy
 import re
 from rag.nlp import bullets_category, is_english, tokenize, remove_contents_table, \
-    hierarchical_merge, make_colon_as_title, naive_merge, random_choices
+    hierarchical_merge, make_colon_as_title, naive_merge, random_choices, tokenize_table
 from rag.nlp import huqie
 from deepdoc.parser import PdfParser, DocxParser
 
     make_colon_as_title(sections)
     bull = bullets_category([t for t in random_choices([t for t,_ in sections], k=100)])
     if bull >= 0: cks = hierarchical_merge(bull, sections, 3)
-    else: cks = naive_merge(sections, kwargs.get("chunk_token_num", 256), kwargs.get("delimer", "\n。；！？"))
+    else:
+        sections = [s.split("@") for s in sections]
+        sections = [(pr[0], "@"+pr[1]) for pr in sections if len(pr)==2]
+        cks = naive_merge(sections, kwargs.get("chunk_token_num", 256), kwargs.get("delimer", "\n。；！？"))
 
-    sections = [t for t, _ in sections]
     # is it English
-    eng = lang.lower() == "english"#is_english(random_choices(sections, k=218))
+    eng = lang.lower() == "english"#is_english(random_choices([t for t, _ in sections], k=218))
+
+    res = tokenize_table(tbls, doc, eng)
 
-    res = []
-    # add tables
-    for img, rows in tbls:
-        bs = 10
-        de = ";" if eng else "；"
-        for i in range(0, len(rows), bs):
-            d = copy.deepcopy(doc)
-            r = de.join(rows[i:i + bs])
-            r = re.sub(r"\t——(来自| in ).*”%s" % de, "", r)
-            tokenize(d, r, eng)
-            d["image"] = img
-            res.append(d)
-            print("TABLE", d["content_with_weight"])
     # wrap up to es documents
     for ck in cks:
         d = copy.deepcopy(doc)

rag/app/manual.py

 import re
 
 from api.db import ParserType
-from rag.nlp import huqie, tokenize
+from rag.nlp import huqie, tokenize, tokenize_table
 from deepdoc.parser import PdfParser
 from rag.utils import num_tokens_from_string
 
     # is it English
     eng = lang.lower() == "english"#pdf_parser.is_english
 
-    res = []
-    # add tables
-    for img, rows in tbls:
-        bs = 10
-        de = ";" if eng else "；"
-        for i in range(0, len(rows), bs):
-            d = copy.deepcopy(doc)
-            r = de.join(rows[i:i + bs])
-            r = re.sub(r"\t——(来自| in ).*”%s" % de, "", r)
-            tokenize(d, r, eng)
-            d["image"] = img
-            res.append(d)
+    res = tokenize_table(tbls, doc, eng)
 
     i = 0
     chunk = []

rag/app/naive.py

 import copy
 import re
 from rag.app import laws
-from rag.nlp import huqie, is_english, tokenize, naive_merge
+from rag.nlp import huqie, is_english, tokenize, naive_merge, tokenize_table
 from deepdoc.parser import PdfParser
 from rag.settings import cron_logger
 
         pdf_parser = Pdf()
         sections, tbls = pdf_parser(filename if not binary else binary,
                               from_page=from_page, to_page=to_page, callback=callback)
-        # add tables
-        for img, rows in tbls:
-            bs = 10
-            de = ";" if eng else "；"
-            for i in range(0, len(rows), bs):
-                d = copy.deepcopy(doc)
-                r = de.join(rows[i:i + bs])
-                r = re.sub(r"\t——(来自| in ).*”%s" % de, "", r)
-                tokenize(d, r, eng)
-                d["image"] = img
-                res.append(d)
+        res = tokenize_table(tbls, doc, eng)
     elif re.search(r"\.txt$", filename, re.IGNORECASE):
         callback(0.1, "Start to parse.")
         txt = ""
     # wrap up to es documents
     for ck in cks:
         print("--", ck)
+        if not ck:continue
         d = copy.deepcopy(doc)
         if pdf_parser:
             d["image"] = pdf_parser.crop(ck)

rag/app/paper.py

 from collections import Counter
 
 from api.db import ParserType
-from rag.nlp import huqie, tokenize
+from rag.nlp import huqie, tokenize, tokenize_table
 from deepdoc.parser import PdfParser
 import numpy as np
 from rag.utils import num_tokens_from_string
     eng = lang.lower() == "english"#pdf_parser.is_english
     print("It's English.....", eng)
 
-    res = []
-    # add tables
-    for img, rows in paper["tables"]:
-        bs = 10
-        de = ";" if eng else "；"
-        for i in range(0, len(rows), bs):
-            d = copy.deepcopy(doc)
-            r = de.join(rows[i:i + bs])
-            r = re.sub(r"\t——(来自| in ).*”%s" % de, "", r)
-            tokenize(d, r)
-            d["image"] = img
-            res.append(d)
+    res = tokenize_table(paper["tables"], doc, eng)
 
     if paper["abstract"]:
         d = copy.deepcopy(doc)

rag/app/presentation.py

 
 class Ppt(PptParser):
     def __call__(self, fnm, from_page, to_page, callback=None):
-        txts = super.__call__(fnm, from_page, to_page)
+        txts = super().__call__(fnm, from_page, to_page)
 
         callback(0.5, "Text extraction finished.")
         import aspose.slides as slides

rag/app/resume.py

     resume = remote_call(filename, binary)
     if len(resume.keys()) < 7:
         callback(-1, "Resume is not successfully parsed.")
-        return []
+        raise Exception("Resume parser remote call fail!")
     callback(0.6, "Done parsing. Chunking...")
     print(json.dumps(resume, ensure_ascii=False, indent=2))
 

rag/nlp/__init__.py

-
+import copy
 
 from nltk.stem import PorterStemmer
 stemmer = PorterStemmer()
         d["content_sm_ltks"] = huqie.qieqie(d["content_ltks"])
 
 
+def tokenize_table(tbls, doc, eng, batch_size=10):
+    res = []
+    # add tables
+    for img, rows in tbls:
+        de = "; " if eng else "； "
+        for i in range(0, len(rows), batch_size):
+            d = copy.deepcopy(doc)
+            r = de.join(rows[i:i + batch_size])
+            tokenize(d, r, eng)
+            d["image"] = img
+            res.append(d)
+    return res
+
+
 def remove_contents_table(sections, eng=False):
     i = 0
     while i < len(sections):
         tnum = num_tokens_from_string(t)
         if tnum < 8: pos = ""
         if tk_nums[-1] > chunk_token_num:
-            cks.append(t + pos)
+            if t.find(pos) < 0: t += pos
+            cks.append(t)
             tk_nums.append(tnum)
         else:
-            cks[-1] += t + pos
+            if cks[-1].find(pos) < 0: t += pos
+            cks[-1] += t
             tk_nums[-1] += tnum
 
     for sec, pos in sections:

rag/nlp/search.py

 # -*- coding: utf-8 -*-
 import json
 import re
+from copy import deepcopy
+
 from elasticsearch_dsl import Q, Search
 from typing import List, Optional, Dict, Union
 from dataclasses import dataclass
                 del s["highlight"]
             q_vec = s["knn"]["query_vector"]
         es_logger.info("【Q】: {}".format(json.dumps(s)))
-        res = self.es.search(s, idxnm=idxnm, timeout="600s", src=src)
+        res = self.es.search(deepcopy(s), idxnm=idxnm, timeout="600s", src=src)
         es_logger.info("TOTAL: {}".format(self.es.getTotal(res)))
         if self.es.getTotal(res) == 0 and "knn" in s:
             bqry, _ = self.qryr.question(qst, min_match="10%")

rag/svr/task_broker.py

                     tsks.append(task)
         else:
             tsks.append(new_task())
-        print(tsks)
+
         bulk_insert_into_db(Task, tsks, True)
         set_dispatching(r["id"])
         tmf.write(str(r["update_time"]) + "\n")

rag/svr/task_executor.py

                             kb_id=row["kb_id"], parser_config=row["parser_config"], tenant_id=row["tenant_id"])
     except Exception as e:
         if re.search("(No such file|not found)", str(e)):
-            callback(-1, "Can not find file <%s>" % row["doc_name"])
+            callback(-1, "Can not find file <%s>" % row["name"])
         else:
             callback(-1, f"Internal server error: %s" %
                      str(e).replace("'", ""))