Mercurial > repos > guerler > springsuite
annotate planemo/lib/python3.7/site-packages/lxml/html/soupparser.py @ 1:56ad4e20f292 draft
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
author | guerler |
---|---|
date | Fri, 31 Jul 2020 00:32:28 -0400 |
parents | |
children |
rev | line source |
---|---|
1
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
1 """External interface to the BeautifulSoup HTML parser. |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
2 """ |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
3 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
4 __all__ = ["fromstring", "parse", "convert_tree"] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
5 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
6 import re |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
7 from lxml import etree, html |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
8 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
9 try: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
10 from bs4 import ( |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
11 BeautifulSoup, Tag, Comment, ProcessingInstruction, NavigableString, |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
12 Declaration, Doctype) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
13 _DECLARATION_OR_DOCTYPE = (Declaration, Doctype) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
14 except ImportError: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
15 from BeautifulSoup import ( |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
16 BeautifulSoup, Tag, Comment, ProcessingInstruction, NavigableString, |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
17 Declaration) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
18 _DECLARATION_OR_DOCTYPE = Declaration |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
19 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
20 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
21 def fromstring(data, beautifulsoup=None, makeelement=None, **bsargs): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
22 """Parse a string of HTML data into an Element tree using the |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
23 BeautifulSoup parser. |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
24 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
25 Returns the root ``<html>`` Element of the tree. |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
26 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
27 You can pass a different BeautifulSoup parser through the |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
28 `beautifulsoup` keyword, and a diffent Element factory function |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
29 through the `makeelement` keyword. By default, the standard |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
30 ``BeautifulSoup`` class and the default factory of `lxml.html` are |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
31 used. |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
32 """ |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
33 return _parse(data, beautifulsoup, makeelement, **bsargs) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
34 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
35 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
36 def parse(file, beautifulsoup=None, makeelement=None, **bsargs): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
37 """Parse a file into an ElemenTree using the BeautifulSoup parser. |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
38 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
39 You can pass a different BeautifulSoup parser through the |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
40 `beautifulsoup` keyword, and a diffent Element factory function |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
41 through the `makeelement` keyword. By default, the standard |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
42 ``BeautifulSoup`` class and the default factory of `lxml.html` are |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
43 used. |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
44 """ |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
45 if not hasattr(file, 'read'): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
46 file = open(file) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
47 root = _parse(file, beautifulsoup, makeelement, **bsargs) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
48 return etree.ElementTree(root) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
49 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
50 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
51 def convert_tree(beautiful_soup_tree, makeelement=None): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
52 """Convert a BeautifulSoup tree to a list of Element trees. |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
53 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
54 Returns a list instead of a single root Element to support |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
55 HTML-like soup with more than one root element. |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
56 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
57 You can pass a different Element factory through the `makeelement` |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
58 keyword. |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
59 """ |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
60 root = _convert_tree(beautiful_soup_tree, makeelement) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
61 children = root.getchildren() |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
62 for child in children: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
63 root.remove(child) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
64 return children |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
65 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
66 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
67 # helpers |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
68 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
69 def _parse(source, beautifulsoup, makeelement, **bsargs): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
70 if beautifulsoup is None: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
71 beautifulsoup = BeautifulSoup |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
72 if hasattr(beautifulsoup, "HTML_ENTITIES"): # bs3 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
73 if 'convertEntities' not in bsargs: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
74 bsargs['convertEntities'] = 'html' |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
75 if hasattr(beautifulsoup, "DEFAULT_BUILDER_FEATURES"): # bs4 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
76 if 'features' not in bsargs: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
77 bsargs['features'] = 'html.parser' # use Python html parser |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
78 tree = beautifulsoup(source, **bsargs) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
79 root = _convert_tree(tree, makeelement) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
80 # from ET: wrap the document in a html root element, if necessary |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
81 if len(root) == 1 and root[0].tag == "html": |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
82 return root[0] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
83 root.tag = "html" |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
84 return root |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
85 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
86 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
87 _parse_doctype_declaration = re.compile( |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
88 r'(?:\s|[<!])*DOCTYPE\s*HTML' |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
89 r'(?:\s+PUBLIC)?(?:\s+(\'[^\']*\'|"[^"]*"))?' |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
90 r'(?:\s+(\'[^\']*\'|"[^"]*"))?', |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
91 re.IGNORECASE).match |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
92 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
93 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
94 class _PseudoTag: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
95 # Minimal imitation of BeautifulSoup.Tag |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
96 def __init__(self, contents): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
97 self.name = 'html' |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
98 self.attrs = [] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
99 self.contents = contents |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
100 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
101 def __iter__(self): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
102 return self.contents.__iter__() |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
103 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
104 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
105 def _convert_tree(beautiful_soup_tree, makeelement): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
106 if makeelement is None: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
107 makeelement = html.html_parser.makeelement |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
108 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
109 # Split the tree into three parts: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
110 # i) everything before the root element: document type |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
111 # declaration, comments, processing instructions, whitespace |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
112 # ii) the root(s), |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
113 # iii) everything after the root: comments, processing |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
114 # instructions, whitespace |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
115 first_element_idx = last_element_idx = None |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
116 html_root = declaration = None |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
117 for i, e in enumerate(beautiful_soup_tree): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
118 if isinstance(e, Tag): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
119 if first_element_idx is None: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
120 first_element_idx = i |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
121 last_element_idx = i |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
122 if html_root is None and e.name and e.name.lower() == 'html': |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
123 html_root = e |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
124 elif declaration is None and isinstance(e, _DECLARATION_OR_DOCTYPE): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
125 declaration = e |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
126 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
127 # For a nice, well-formatted document, the variable roots below is |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
128 # a list consisting of a single <html> element. However, the document |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
129 # may be a soup like '<meta><head><title>Hello</head><body>Hi |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
130 # all<\p>'. In this example roots is a list containing meta, head |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
131 # and body elements. |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
132 if first_element_idx is None: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
133 pre_root = post_root = [] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
134 roots = beautiful_soup_tree.contents |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
135 else: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
136 pre_root = beautiful_soup_tree.contents[:first_element_idx] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
137 roots = beautiful_soup_tree.contents[first_element_idx:last_element_idx+1] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
138 post_root = beautiful_soup_tree.contents[last_element_idx+1:] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
139 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
140 # Reorganize so that there is one <html> root... |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
141 if html_root is not None: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
142 # ... use existing one if possible, ... |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
143 i = roots.index(html_root) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
144 html_root.contents = roots[:i] + html_root.contents + roots[i+1:] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
145 else: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
146 # ... otherwise create a new one. |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
147 html_root = _PseudoTag(roots) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
148 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
149 convert_node = _init_node_converters(makeelement) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
150 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
151 # Process pre_root |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
152 res_root = convert_node(html_root) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
153 prev = res_root |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
154 for e in reversed(pre_root): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
155 converted = convert_node(e) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
156 if converted is not None: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
157 prev.addprevious(converted) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
158 prev = converted |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
159 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
160 # ditto for post_root |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
161 prev = res_root |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
162 for e in post_root: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
163 converted = convert_node(e) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
164 if converted is not None: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
165 prev.addnext(converted) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
166 prev = converted |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
167 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
168 if declaration is not None: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
169 try: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
170 # bs4 provides full Doctype string |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
171 doctype_string = declaration.output_ready() |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
172 except AttributeError: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
173 doctype_string = declaration.string |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
174 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
175 match = _parse_doctype_declaration(doctype_string) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
176 if not match: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
177 # Something is wrong if we end up in here. Since soupparser should |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
178 # tolerate errors, do not raise Exception, just let it pass. |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
179 pass |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
180 else: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
181 external_id, sys_uri = match.groups() |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
182 docinfo = res_root.getroottree().docinfo |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
183 # strip quotes and update DOCTYPE values (any of None, '', '...') |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
184 docinfo.public_id = external_id and external_id[1:-1] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
185 docinfo.system_url = sys_uri and sys_uri[1:-1] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
186 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
187 return res_root |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
188 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
189 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
190 def _init_node_converters(makeelement): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
191 converters = {} |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
192 ordered_node_types = [] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
193 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
194 def converter(*types): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
195 def add(handler): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
196 for t in types: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
197 converters[t] = handler |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
198 ordered_node_types.append(t) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
199 return handler |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
200 return add |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
201 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
202 def find_best_converter(node): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
203 for t in ordered_node_types: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
204 if isinstance(node, t): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
205 return converters[t] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
206 return None |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
207 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
208 def convert_node(bs_node, parent=None): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
209 # duplicated in convert_tag() below |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
210 try: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
211 handler = converters[type(bs_node)] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
212 except KeyError: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
213 handler = converters[type(bs_node)] = find_best_converter(bs_node) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
214 if handler is None: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
215 return None |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
216 return handler(bs_node, parent) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
217 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
218 def map_attrs(bs_attrs): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
219 if isinstance(bs_attrs, dict): # bs4 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
220 attribs = {} |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
221 for k, v in bs_attrs.items(): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
222 if isinstance(v, list): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
223 v = " ".join(v) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
224 attribs[k] = unescape(v) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
225 else: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
226 attribs = dict((k, unescape(v)) for k, v in bs_attrs) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
227 return attribs |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
228 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
229 def append_text(parent, text): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
230 if len(parent) == 0: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
231 parent.text = (parent.text or '') + text |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
232 else: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
233 parent[-1].tail = (parent[-1].tail or '') + text |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
234 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
235 # converters are tried in order of their definition |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
236 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
237 @converter(Tag, _PseudoTag) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
238 def convert_tag(bs_node, parent): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
239 attrs = bs_node.attrs |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
240 if parent is not None: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
241 attribs = map_attrs(attrs) if attrs else None |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
242 res = etree.SubElement(parent, bs_node.name, attrib=attribs) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
243 else: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
244 attribs = map_attrs(attrs) if attrs else {} |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
245 res = makeelement(bs_node.name, attrib=attribs) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
246 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
247 for child in bs_node: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
248 # avoid double recursion by inlining convert_node(), see above |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
249 try: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
250 handler = converters[type(child)] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
251 except KeyError: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
252 pass |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
253 else: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
254 if handler is not None: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
255 handler(child, res) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
256 continue |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
257 convert_node(child, res) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
258 return res |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
259 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
260 @converter(Comment) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
261 def convert_comment(bs_node, parent): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
262 res = html.HtmlComment(bs_node) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
263 if parent is not None: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
264 parent.append(res) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
265 return res |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
266 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
267 @converter(ProcessingInstruction) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
268 def convert_pi(bs_node, parent): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
269 if bs_node.endswith('?'): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
270 # The PI is of XML style (<?as df?>) but BeautifulSoup |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
271 # interpreted it as being SGML style (<?as df>). Fix. |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
272 bs_node = bs_node[:-1] |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
273 res = etree.ProcessingInstruction(*bs_node.split(' ', 1)) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
274 if parent is not None: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
275 parent.append(res) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
276 return res |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
277 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
278 @converter(NavigableString) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
279 def convert_text(bs_node, parent): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
280 if parent is not None: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
281 append_text(parent, unescape(bs_node)) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
282 return None |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
283 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
284 return convert_node |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
285 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
286 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
287 # copied from ET's ElementSoup |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
288 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
289 try: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
290 from html.entities import name2codepoint # Python 3 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
291 except ImportError: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
292 from htmlentitydefs import name2codepoint |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
293 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
294 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
295 handle_entities = re.compile(r"&(\w+);").sub |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
296 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
297 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
298 try: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
299 unichr |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
300 except NameError: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
301 # Python 3 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
302 unichr = chr |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
303 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
304 |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
305 def unescape(string): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
306 if not string: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
307 return '' |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
308 # work around oddities in BeautifulSoup's entity handling |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
309 def unescape_entity(m): |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
310 try: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
311 return unichr(name2codepoint[m.group(1)]) |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
312 except KeyError: |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
313 return m.group(0) # use as is |
56ad4e20f292
"planemo upload commit 6eee67778febed82ddd413c3ca40b3183a3898f1"
guerler
parents:
diff
changeset
|
314 return handle_entities(unescape_entity, string) |