lxbfYeaa/BookConverter
Archived
Template
1
0
Fork 0
forked from LiveCarta/BookConverter
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge pull request #307 from Teqniksoft/kiryl/converter_fix

Browse Source 
Kiryl/converter fix
This commit is contained in:
bivis
2022-10-14 10:27:39 +03:00
committed by GitHub
parent bb2e46062e 1363940fa2
commit 7169256a41
5 changed files with 144 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files

86
presets/docx_presets.json
View File

@@ -12,9 +12,13 @@
"name": "id", "name": "id",
"value": "^Table of Contents\\d+" "value": "^Table of Contents\\d+"
} }
] ],
"text": null
}, },
"tag_to_wrap": "TOC" "tag_to_wrap": {
"name": "TOC",
"attrs": []
}
} }
] ]
}, },
@@ -35,7 +39,8 @@
"name": "id", "name": "id",
"value": "^Table of Contents\\d+" "value": "^Table of Contents\\d+"
} }
] ],
"text": null
} }
} }
] ]
@@ -46,7 +51,10 @@
{ {
"tags": ["^h[6-9]$"], "tags": ["^h[6-9]$"],
"condition": null, "condition": null,
"tag_to_replace": "p" "tag_to_replace": {
"name": "p",
"attrs": null
}
}, },
{ {
"tags": ["^div$"], "tags": ["^div$"],
@@ -58,9 +66,13 @@
"name": "style", "name": "style",
"value": "column-count: 2" "value": "column-count: 2"
} }
] ],
"text": null
}, },
"tag_to_replace": "p" "tag_to_replace": {
"name": "p",
"attrs": null
}
} }
] ]
}, },
@@ -68,12 +80,14 @@
"preset_name": "attr_replacer", "preset_name": "attr_replacer",
"rules": [ "rules": [
{ {
"attr": { "tags": ["^p$"],
"name": "style",
"value": "column-count: 2"
},
"condition": { "condition": {
"tags": ["^p$"] "attrs": [
{
"name": "style",
"value": "column-count: 2"
}
]
}, },
"attr_to_replace": { "attr_to_replace": {
"name": "class", "name": "class",
@@ -103,7 +117,8 @@
"name": "face", "name": "face",
"value": "^Times New Roman[\\w, ]+$" "value": "^Times New Roman[\\w, ]+$"
} }
] ],
"text": null
} }
}, },
{ {
@@ -111,7 +126,8 @@
"condition": { "condition": {
"parent_tags": ":is(li)", "parent_tags": ":is(li)",
"child_tags": null, "child_tags": null,
"attrs": null "attrs": null,
"text": null
} }
}, },
{ {
@@ -124,7 +140,8 @@
"name": "name", "name": "name",
"value": "_GoBack" "value": "_GoBack"
} }
] ],
"text": null
} }
}, },
{ {
@@ -132,7 +149,8 @@
"condition": { "condition": {
"parent_tags": ":is(a)", "parent_tags": ":is(a)",
"child_tags": ":is(a)", "child_tags": ":is(a)",
"attrs": null "attrs": null,
"text": null
} }
}, },
{ {
@@ -140,7 +158,8 @@
"condition": { "condition": {
"parent_tags": ":is(h1, h2, h3, h4, h5, h6, h7, h8, h9)", "parent_tags": ":is(h1, h2, h3, h4, h5, h6, h7, h8, h9)",
"child_tags": null, "child_tags": null,
"attrs": null "attrs": null,
"text": null
} }
}, },
{ {
@@ -148,5 +167,40 @@
"condition": null "condition": null
} }
] ]
},
{
"preset_name": "inserter",
"rules": [
{
"tags": ["^p$"],
"condition": {
"parent_tags": null,
"child_tags": null,
"attrs": null,
"text": "\\$\\$[\\s\\S]*?\\$\\$"
},
"tag_to_insert": {
"name": "span",
"attrs": [
{
"name": "class",
"value": "math-tex"
}
]
}
}
]
},
{
"preset_name": "text_replacer",
"rules": [
{
"tags": ["^p$"],
"condition": {
"text": "(\\\\nonumber\\\\\\\\\\\\noalign{\\\\pagebreak}[\\s\\S]*?)\\\\"
},
"text_to_replace": "\\\\"
}
]
} }
] ]

50
presets/epub_presets.json
View File

@@ -48,17 +48,23 @@
{ {
"tags": ["^h[6-9]$", "^figure$", "^section$", "^div$", "blockquote"], "tags": ["^h[6-9]$", "^figure$", "^section$", "^div$", "blockquote"],
"condition": null, "condition": null,
"tag_to_replace": "p" "tag_to_replace": {
"name": "p"
}
}, },
{ {
"tags": ["^aside$"], "tags": ["^aside$"],
"condition": null, "condition": null,
"tag_to_replace": "div" "tag_to_replace": {
"name": "div"
}
}, },
{ {
"tags": ["^header$", "^footer$"], "tags": ["^header$", "^footer$"],
"condition": null, "condition": null,
"tag_to_replace": "span" "tag_to_replace": {
"name": "span"
}
}, },
{ {
"tags": ["^code$", "^kbd$", "^var$"], "tags": ["^code$", "^kbd$", "^var$"],
@@ -67,22 +73,30 @@
"child_tags": null, "child_tags": null,
"attrs": null "attrs": null
}, },
"tag_to_replace": "span" "tag_to_replace": {
"name": "span"
}
}, },
{ {
"tags": ["^em$"], "tags": ["^em$"],
"condition": null, "condition": null,
"tag_to_replace": "i" "tag_to_replace": {
"name": "i"
}
}, },
{ {
"tags": ["^b$"], "tags": ["^b$"],
"condition": null, "condition": null,
"tag_to_replace": "strong" "tag_to_replace": {
"name": "strong"
}
}, },
{ {
"tags": ["^image$"], "tags": ["^image$"],
"condition": null, "condition": null,
"tag_to_replace": "img" "tag_to_replace": {
"name": "img"
}
} }
] ]
}, },
@@ -90,12 +104,14 @@
"preset_name": "attr_replacer", "preset_name": "attr_replacer",
"rules": [ "rules": [
{ {
"attr": { "tags": ["^img$"],
"name": "xlink:href",
"value": ".*"
},
"condition": { "condition": {
"tags": ["^img$"] "attrs": [
{
"name": "xlink:href",
"value": ".*"
}
]
}, },
"attr_to_replace": { "attr_to_replace": {
"name": "src", "name": "src",
@@ -140,12 +156,18 @@
"child_tags": ":not(:has(code, kbd, var))", "child_tags": ":not(:has(code, kbd, var))",
"attrs": null "attrs": null
}, },
"tag_to_insert": "code" "tag_to_insert": {
"name": "code",
"attrs": []
}
}, },
{ {
"tags": ["^h[1-5]$"], "tags": ["^h[1-5]$"],
"condition": null, "condition": null,
"tag_to_insert": "strong" "tag_to_insert": {
"name":"strong",
"attrs": []
}
} }
] ]
} }

4
src/docx_converter/docx_solver.py
View File

@@ -80,7 +80,8 @@ class DocxBook(BookSolver):
if __name__ == "__main__": if __name__ == "__main__":
docx_file_path = "../../books/docx/AmericanGovernment3e-WEB.docx"
docx_file_path = "../../books/docx/output.docx"
logger_object = BookLogger( logger_object = BookLogger(
name="docx", book_id=docx_file_path.split("/")[-1]) name="docx", book_id=docx_file_path.split("/")[-1])
locker = Event() locker = Event()
@@ -88,7 +89,6 @@ if __name__ == "__main__":
html_converter = Docx2LibreHtml(file_path=docx_file_path, html_converter = Docx2LibreHtml(file_path=docx_file_path,
logger=logger_object, libre_locker=locker) logger=logger_object, libre_locker=locker)
html_preprocessor = HtmlPresetsProcessor( html_preprocessor = HtmlPresetsProcessor(
logger=logger_object, preset_path="../../presets/docx_presets.json") logger=logger_object, preset_path="../../presets/docx_presets.json")
style_preprocessor = StyleReader() style_preprocessor = StyleReader()

2
src/epub_converter/epub_solver.py
View File

@@ -43,7 +43,7 @@ class EpubBook(BookSolver):
if __name__ == "__main__": if __name__ == "__main__":
epub_file_path = "../../books/epub/9780763774134.epub" epub_file_path = "../../books/epub/9781634259804.epub"
logger_object = BookLogger( logger_object = BookLogger(
name="epub", book_id=epub_file_path.split("/")[-1]) name="epub", book_id=epub_file_path.split("/")[-1])

51
src/html_presets_processor.py
View File

@@ -18,13 +18,14 @@ class HtmlPresetsProcessor:
"replacer": self._replace_tag, "replacer": self._replace_tag,
"attr_replacer": self._replace_attr, "attr_replacer": self._replace_attr,
"unwrapper": self._unwrap_tag, "unwrapper": self._unwrap_tag,
"inserter": self._insert_tag "inserter": self._insert_tag,
"text_replacer": self._replace_text
} }
@staticmethod @staticmethod
def _wrap_tag(**kwargs): def _wrap_tag(**kwargs):
kwargs["tag"].wrap(kwargs["body_tag"].new_tag( kwargs["tag"].wrap(kwargs["body_tag"].new_tag(
kwargs["rule"]["tag_to_wrap"])) kwargs["rule"]["tag_to_wrap"]["name"]))
@staticmethod @staticmethod
def set_attrs_to_parent(tag, parent_tag): def set_attrs_to_parent(tag, parent_tag):
@@ -98,24 +99,31 @@ class HtmlPresetsProcessor:
@staticmethod @staticmethod
def _replace_tag(**kwargs): def _replace_tag(**kwargs):
tag_to_replace: str = kwargs["rule"]["tag_to_replace"] tag_to_replace: str = kwargs["rule"]["tag_to_replace"]["name"]
kwargs["tag"].name = tag_to_replace kwargs["tag"].name = tag_to_replace
if kwargs["rule"]["tag_to_replace"].get("attrs"):
dict_attributes = {attr["name"]: attr["value"]
for attr in kwargs["rule"]["tag_to_replace"]["attrs"]}
kwargs["tag"].attrs = dict_attributes
@staticmethod @staticmethod
def _replace_attr(**kwargs): def _replace_attr(**kwargs):
attr, attr_value =\ attr = kwargs["rule"]["condition"]["attrs"][0]
kwargs["rule"]["attr"]["name"], kwargs["rule"]["attr"]["value"] attr_name, attr_value =\
attr["name"], attr["value"]
attr_to_replace, attr_value_to_replace =\ attr_to_replace, attr_value_to_replace =\
kwargs["rule"]["attr_to_replace"]["name"], kwargs["rule"]["attr_to_replace"]["value"] kwargs["rule"]["attr_to_replace"]["name"], kwargs["rule"]["attr_to_replace"]["value"]
if attr_to_replace: if attr_to_replace:
kwargs["tag"][attr_to_replace] = kwargs["tag"][attr] kwargs["tag"][attr_to_replace] = kwargs["tag"][attr_name] \
if kwargs["tag"].get(attr_name)\
else ""
if attr_value_to_replace: if attr_value_to_replace:
kwargs["tag"].attrs[attr_to_replace] = attr_value_to_replace kwargs["tag"].attrs[attr_to_replace] = attr_value_to_replace
del kwargs["tag"][attr] del kwargs["tag"][attr_name]
elif attr_value_to_replace: elif attr_value_to_replace:
kwargs["tag"].attrs[attr] = attr_value_to_replace kwargs["tag"].attrs[attr_name] = attr_value_to_replace
elif attr: elif attr_name:
del kwargs["tag"][attr] del kwargs["tag"][attr_name]
def _unwrap_tag(self, **kwargs): def _unwrap_tag(self, **kwargs):
if kwargs["tag"].parent: if kwargs["tag"].parent:
@@ -124,14 +132,24 @@ class HtmlPresetsProcessor:
@staticmethod @staticmethod
def _insert_tag(**kwargs): def _insert_tag(**kwargs):
dict_attributes = {attr["name"]: attr["value"]
for attr in kwargs["rule"]["tag_to_insert"]["attrs"]}
tag_to_insert = \ tag_to_insert = \
kwargs["body_tag"].new_tag(kwargs["rule"]["tag_to_insert"]) kwargs["body_tag"].new_tag(
kwargs["rule"]["tag_to_insert"]["name"], attrs=dict_attributes)
# insert all items that was in tag to subtag and remove from tag # insert all items that was in tag to subtag and remove from tag
for content in reversed(kwargs["tag"].contents): for content in reversed(kwargs["tag"].contents):
tag_to_insert.insert(0, content.extract()) tag_to_insert.insert(0, content.extract())
# wrap subtag with items # wrap subtag with items
kwargs["tag"].append(tag_to_insert) kwargs["tag"].append(tag_to_insert)
@staticmethod
def _replace_text(**kwargs):
if re.search(re.compile(kwargs["rule"]["condition"]["text"]), kwargs["tag"].string):
new_text = re.sub(re.compile(
kwargs["rule"]["condition"]["text"]), kwargs["rule"]["text_to_replace"], kwargs["tag"].string)
kwargs["tag"].string.replace_with(new_text)
@staticmethod @staticmethod
def _process_tags(body_tag: BeautifulSoup, def _process_tags(body_tag: BeautifulSoup,
rules: List[Dict[str, Union[List[str], str, Dict[str, Union[List[Dict[str, str]], int, str]]]]], rules: List[Dict[str, Union[List[str], str, Dict[str, Union[List[Dict[str, str]], int, str]]]]],
@@ -171,11 +189,12 @@ class HtmlPresetsProcessor:
for tag in body_tag.find_all([re.compile(tag) for tag in tags], for tag in body_tag.find_all([re.compile(tag) for tag in tags],
{attr["name"]: re.compile(fr"{attr['value']}")}): {attr["name"]: re.compile(fr"{attr['value']}")}):
action(body_tag=body_tag, tag=tag, rule=rule) action(body_tag=body_tag, tag=tag, rule=rule)
# attr replacer elif condition_on_tag[0] == "text":
elif condition_on_tag[0] == "tags": # find all tags that are in List of tags and tags that contains required text
attr = rule["attr"] for tag in body_tag.find_all(
for tag in body_tag.find_all([re.compile(tag) for tag in tags], lambda t: re.search(r"(?=(" + '|'.join([tag for tag in tags]) + r"))",
{attr['name']: re.compile(fr"{attr['value']}")}): t.name) and re.search(re.compile(rule["condition"]["text"]),
t.text)):
action(body_tag=body_tag, tag=tag, rule=rule) action(body_tag=body_tag, tag=tag, rule=rule)
else: else:
for tag in body_tag.find_all([re.compile(tag) for tag in tags]): for tag in body_tag.find_all([re.compile(tag) for tag in tags]):