From 9a843667b3e4fb8ee4ebc9d5010b00b41f770370 Mon Sep 17 00:00:00 2001
From: KevinHuSh <kevinhu.sh@gmail.com>
Date: Tue, 19 Mar 2024 15:31:47 +0800
Subject: [PATCH] fix github account login issue (#132)

---
 api/apps/user_app.py         |  2 ++
 deepdoc/parser/pdf_parser.py |  2 +-
 rag/app/manual.py            |  5 +----
 rag/app/qa.py                | 12 +++++++++---
 rag/nlp/__init__.py          |  1 +
 rag/nlp/query.py             |  2 +-
 6 files changed, 15 insertions(+), 9 deletions(-)

diff --git a/api/apps/user_app.py b/api/apps/user_app.py
index 7ff2364f6..c3dd68ef2 100644
--- a/api/apps/user_app.py
+++ b/api/apps/user_app.py
@@ -106,7 +106,9 @@ def github_callback():
             stat_logger.exception(e)
             return redirect("/?error=%s"%str(e))
     user = users[0]
+    user.access_token = get_uuid()
     login_user(user)
+    user.save()
     return redirect("/?auth=%s" % user.get_id())
 
 
diff --git a/deepdoc/parser/pdf_parser.py b/deepdoc/parser/pdf_parser.py
index 8aaa7dc51..2f6bec07b 100644
--- a/deepdoc/parser/pdf_parser.py
+++ b/deepdoc/parser/pdf_parser.py
@@ -639,7 +639,7 @@ class HuParser:
                 mink = ""
                 minv = 1000000000
                 for k, bxs in tbls.items():
-                    for b in bxs[:10]:
+                    for b in bxs:
                         if b.get("layout_type", "").find("caption") >= 0:
                             continue
                         y_dis = self._y_dis(c, b)
diff --git a/rag/app/manual.py b/rag/app/manual.py
index 018a3b725..b3dc688ec 100644
--- a/rag/app/manual.py
+++ b/rag/app/manual.py
@@ -62,9 +62,6 @@ class Pdf(PdfParser):
         for b in self.boxes:
             b["text"] = re.sub(r"([\t 　]|\u3000){2,}", " ", b["text"].strip())
 
-        # merge chunks with the same bullets
-        self._merge_with_same_bullet()
-
         # set pivot using the most frequent type of title,
         # then merge between 2 pivot
         bull = bullets_category([b["text"] for b in self.boxes])
@@ -79,7 +76,7 @@ class Pdf(PdfParser):
 
         sections = [(b["text"], sec_ids[i], get_position(b)) for i, b in enumerate(self.boxes)]
         for (img, rows), poss in tbls:
-            sections.append((rows[0], -1, [(p[0]+1, p[1], p[2], p[3], p[4]) for p in poss]))
+            sections.append((rows if isinstance(rows, str) else rows[0], -1, [(p[0]+1-from_page, p[1], p[2], p[3], p[4]) for p in poss]))
 
         chunks = []
         last_sid = -2
diff --git a/rag/app/qa.py b/rag/app/qa.py
index e649b4a23..0a546921e 100644
--- a/rag/app/qa.py
+++ b/rag/app/qa.py
@@ -11,6 +11,7 @@
 #  limitations under the License.
 #
 import re
+from copy import deepcopy
 from io import BytesIO
 from nltk import word_tokenize
 from openpyxl import load_workbook
@@ -93,12 +94,17 @@ def chunk(filename, binary=None, lang="Chinese", callback=None, **kwargs):
         All the deformed lines will be ignored.
         Every pair of Q&A will be treated as a chunk.
     """
+    eng = lang.lower() == "english"
     res = []
+    doc = {
+        "docnm_kwd": filename,
+        "title_tks": huqie.qie(re.sub(r"\.[a-zA-Z]+$", "", filename))
+    }
     if re.search(r"\.xlsx?$", filename, re.IGNORECASE):
         callback(0.1, "Start to parse.")
         excel_parser = Excel()
         for q, a in excel_parser(filename, binary, callback):
-            res.append(beAdoc({}, q, a, excel_parser.is_english))
+            res.append(beAdoc(deepcopy(doc), q, a, eng))
         return res
     elif re.search(r"\.(txt|csv)$", filename, re.IGNORECASE):
         callback(0.1, "Start to parse.")
@@ -113,14 +119,14 @@ def chunk(filename, binary=None, lang="Chinese", callback=None, **kwargs):
                         break
                     txt += l
         lines = txt.split("\n")
-        eng = lang.lower() == "english"#is_english([rmPrefix(l) for l in lines[:100]])
+        #is_english([rmPrefix(l) for l in lines[:100]])
         fails = []
         for i, line in enumerate(lines):
             arr = [l for l in line.split("\t") if len(l) > 1]
             if len(arr) != 2:
                 fails.append(str(i))
                 continue
-            res.append(beAdoc({}, arr[0], arr[1], eng))
+            res.append(beAdoc(deepcopy(doc), arr[0], arr[1], eng))
             if len(res) % 999 == 0:
                 callback(len(res) * 0.6 / len(lines), ("Extract Q&A: {}".format(len(res)) + (
                     f"{len(fails)} failure, line: %s..." % (",".join(fails[:3])) if fails else "")))
diff --git a/rag/nlp/__init__.py b/rag/nlp/__init__.py
index 95e8308ea..4afd37624 100644
--- a/rag/nlp/__init__.py
+++ b/rag/nlp/__init__.py
@@ -76,6 +76,7 @@ def is_english(texts):
 
 def tokenize(d, t, eng):
     d["content_with_weight"] = t
+    t = re.sub(r"</?(table|td|caption|tr|th)( [^<>]{0,12})?>", " ", t)
     if eng:
         t = re.sub(r"([a-z])-([a-z])", r"\1\2", t)
         d["content_ltks"] = " ".join([stemmer.stem(w)
diff --git a/rag/nlp/query.py b/rag/nlp/query.py
index 3d944b3be..608f18f99 100644
--- a/rag/nlp/query.py
+++ b/rag/nlp/query.py
@@ -29,7 +29,7 @@ class EsQueryer:
         for t in arr:
             if not re.match(r"[a-zA-Z]+$", t):
                 e += 1
-        return e * 1. / len(arr) >= 0.8
+        return e * 1. / len(arr) >= 0.7
 
     @staticmethod
     def rmWWW(txt):