datamule 1.3.1__py3-none-any.whl → 1.4.2__py3-none-any.whl

datamule/document/document.py

@@ -3,11 +3,14 @@ import csv
 import re
 from doc2dict import xml2dict, txt2dict, dict2dict
 from doc2dict.mapping import flatten_hierarchy
+from doc2dict import html2dict, visualize_dict, get_title, unnest_dict, pdf2dict
 from ..mapping_dicts.txt_mapping_dicts import dict_10k, dict_10q, dict_8k, dict_13d, dict_13g
 from ..mapping_dicts.xml_mapping_dicts import dict_345
+from ..mapping_dicts.html_mapping_dicts import dict_10k_html, dict_10q_html, dict_8k_html
 from selectolax.parser import HTMLParser
 from .processing import process_tabular_data
 from pathlib import Path
+import webbrowser
 
 class Document:
     def __init__(self, type, content, extension,accession,filing_date,path=None):
@@ -99,26 +102,10 @@ class Document:
         if self.data:
             return self.data
         
-        # preprocess content
-        if self.extension == '.txt':
-            self.content = self._preprocess_txt_content()
-        elif self.extension in ['.htm', '.html']:
-            self.content = self._preprocess_html_content()
-
         mapping_dict = None
-
-        if self.extension == '.xml':
-            if self.type in ['3', '4', '5', '3/A', '4/A', '5/A']:
-                mapping_dict = dict_345
-
-            self.data = xml2dict(content=self.content, mapping_dict=mapping_dict)
-
-
-
-        # will deprecate this when we add html2dict
-        elif self.extension in ['.htm', '.html','.txt']:
-
-
+        
+        if self.extension == '.txt':
+            content = self._preprocess_txt_content()
             if self.type == '10-Q':
                 mapping_dict = dict_10q
             elif self.type == '10-K':
@@ -131,8 +118,26 @@ class Document:
                 mapping_dict = dict_13g
             
             self.data = {}
-            self.data['document'] = dict2dict(txt2dict(content=self.content, mapping_dict=mapping_dict))
-        return self.data
+            self.data['document'] = dict2dict(txt2dict(content=content, mapping_dict=mapping_dict))
+        elif self.extension in ['.htm', '.html']:
+            if self.type == '10-K':
+                mapping_dict = dict_10k_html
+            elif self.type == '10-Q':
+                mapping_dict = dict_10q_html
+            elif self.type == '8-K':
+                mapping_dict = dict_8k_html
+            
+            dct = html2dict(content=self.content, mapping_dict=mapping_dict)
+            self.data = dct
+        elif self.extension == '.xml':
+            if self.type in ['3', '4', '5', '3/A', '4/A', '5/A']:
+                mapping_dict = dict_345
+            
+            self.data = xml2dict(content=self.content, mapping_dict=mapping_dict)
+        elif self.extension == '.pdf':
+            self.data = pdf2dict(content=self.content, mapping_dict=mapping_dict)
+        else:
+            pass
     
     def write_json(self, output_filename=None):
         if not self.data:
@@ -206,9 +211,33 @@ class Document:
                     })
         
         return items
+    
+    def visualize(self):
+        if not self.data:
+            self.parse()
+
+        if not self.data:
+            if self.extension in ['.jpg', '.png', '.pdf']:
+                webbrowser.open('file://' + str(self.path))
+            else:
+                pass
+        else:
+            visualize_dict(self.data)
+
+    def get_section(self, title, format='dict'):
+        if not self.data:
+            self.parse()
+
+        result = get_title(self.data,title)
+
+        if format == 'text':
+            result = [item[1] for item in result]
+            result = [unnest_dict(item) for item in result]
+
+        return result
+
    
-   # this will all have to be changed. default will be to flatten everything
-   # candidate for deletion
+   # TODO CHANGE THIS
     def __iter__(self):
         self.parse()
 

datamule/mapping_dicts/html_mapping_dicts.py

@@ -0,0 +1,11 @@
+dict_10k_html = {
+    ('part',r'^part\s*([ivx]+)$') : 0,
+    ('signatures',r'^signatures?\.*$') : 0,
+    ('item',r'^item\s*(\d+)\.?([a-z])?') : 1,
+}
+dict_10q_html = dict_10k_html
+
+dict_8k_html = {
+    ('signatures',r'^signatures?\.*$') : 0,
+    ('item',r'^item\s*(\d+\.\d+)') : 0,
+}

{datamule-1.3.1.dist-info → datamule-1.4.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: datamule
-Version: 1.3.1
+Version: 1.4.2
 Summary: Work with SEC submissions at scale.
 Home-page: https://github.com/john-friedman/datamule-python
 Author: John Friedman

{datamule-1.3.1.dist-info → datamule-1.4.2.dist-info}/RECORD

@@ -8,7 +8,7 @@ datamule/sheet.py,sha256=TvFqK9eAYuVoJ2uWdAlx5EN6vS9lke-aZf7FqtUiDBc,22304
 datamule/submission.py,sha256=Yh5nG3ioumhl6z30wJdIEmKjDDNSuo0r2xycZSIaeIg,11035
 datamule/data/listed_filer_metadata.csv,sha256=dT9fQ8AC5P1-Udf_UF0ZkdXJ88jNxJb_tuhi5YYL1rc,2426827
 datamule/document/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-datamule/document/document.py,sha256=menUFoeWwiY0rJnBkQiqY4NWnO0J17-qs8jFvO_1jiY,9969
+datamule/document/document.py,sha256=uHpkLyl6er4MdZDDJNzxSdtwkfDMtMVeG01uRCRblfA,11002
 datamule/document/processing.py,sha256=jDCEzBFDSQtq7nQxRScIsbALnFcvMPOkNkMUCa7mFxg,31921
 datamule/document/table.py,sha256=73yUJKY82ap32jhLmZeTti-jQ_lyhcJGlGwyxLtgYOg,12944
 datamule/document/mappings/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -36,6 +36,7 @@ datamule/document/mappings/thirteenfhr.py,sha256=XpYRIMPZnGLfEE4TqBI0BPXbyuq0xf3
 datamule/document/mappings/twentyfivense.py,sha256=lKyj0ZBhkHX9gQJMTUPrQlxYFg3k-aBnWqtoS5bujZM,905
 datamule/document/mappings/twentyfourf2nt.py,sha256=Q7RPT3JgJHjYdjMuaSyAxclt6QPT_LgCQloxp-ByDuI,4118
 datamule/mapping_dicts/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+datamule/mapping_dicts/html_mapping_dicts.py,sha256=Tfg7wX6EUUtsg_HD2KHH3Z6S3GR9FZc1pwL8kCGfRRU,282
 datamule/mapping_dicts/txt_mapping_dicts.py,sha256=DQPrGYbAPQxomRUtt4iiMGrwuF7BHc_LeFBQuYBzU9o,6311
 datamule/mapping_dicts/xml_mapping_dicts.py,sha256=Z22yDVwKYonUfM5foQP00dVDE8EHhhMKp0CLqVKV5OI,438
 datamule/sec/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -57,7 +58,7 @@ datamule/seclibrary/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSu
 datamule/seclibrary/bq.py,sha256=C8sb_rpXTvchprrFLcbRar4Qi0XWW25tnv1YsHSS5o4,18025
 datamule/seclibrary/downloader.py,sha256=PIgz_7ASUTZOHcUZGcD1SmLaGSbq7xe7EiJT0Z7HU4M,13653
 datamule/seclibrary/query.py,sha256=qGuursTERRbOGfoDcYcpo4oWkW3PCBW6x1Qf1Puiak4,7352
-datamule-1.3.1.dist-info/METADATA,sha256=rh8_hyfGhdxK7Ip1-LKomXj4c-iu2eGRX0NfyjEGOnk,469
-datamule-1.3.1.dist-info/WHEEL,sha256=oiQVh_5PnQM0E3gPdiz09WCNmwiHDMaGer_elqB3coM,92
-datamule-1.3.1.dist-info/top_level.txt,sha256=iOfgmtSMFVyr7JGl_bYSTDry79JbmsG4p8zKq89ktKk,9
-datamule-1.3.1.dist-info/RECORD,,
+datamule-1.4.2.dist-info/METADATA,sha256=mUujGnhNFWQFCH8qg2aJ4_bS4Zp-XhEkD_Y6x9qAh9Q,469
+datamule-1.4.2.dist-info/WHEEL,sha256=oiQVh_5PnQM0E3gPdiz09WCNmwiHDMaGer_elqB3coM,92
+datamule-1.4.2.dist-info/top_level.txt,sha256=iOfgmtSMFVyr7JGl_bYSTDry79JbmsG4p8zKq89ktKk,9
+datamule-1.4.2.dist-info/RECORD,,