html.diff

Module Contents

Classes

DEL_START()
DEL_END()
NoDeletes() Raised when the document no longer contains any pending deletes
token() Represents a diffable token, generally a word that is displayed to
tag_token() Represents a token that is actually a tag. Currently this is just
href_token() Represents the href in an anchor tag. Unlike other words, we only
InsensitiveSequenceMatcher() Acts like SequenceMatcher, but tries not to find very small equal

Functions

default_markup(text,version)
html_annotate(doclist,markup=default_markup) doclist should be ordered from oldest to newest, like:
tokenize_annotated(doc,annotation) Tokenize a document and add an annotation attribute to each token
html_annotate_merge_annotations(tokens_old,tokens_new) Merge the annotations from tokens_old into tokens_new, when the
copy_annotations(src,dest) Copy annotations from the tokens listed in src to the tokens in dest
compress_tokens(tokens) Combine adjacent tokens when there is no HTML between the tokens,
compress_merge_back(tokens,tok) Merge tok into the last element of tokens (modifying the list of
markup_serialize_tokens(tokens,markup_func) Serialize the list of tokens into a list of text chunks, calling
htmldiff(old_html,new_html) Do a diff of the old and new document. The documents are HTML
htmldiff_tokens(html1_tokens,html2_tokens) Does a diff on the tokens themselves, returning a list of text
expand_tokens(tokens,equal=False) Given a list of tokens, return a generator of the chunks of
merge_insert(ins_chunks,doc) doc is the already-handled document (as a list of text chunks);
merge_delete(del_chunks,doc) Adds the text chunks in del_chunks to the document doc (another
cleanup_delete(chunks) Cleans up any DEL_START/DEL_END markers in the document, replacing
split_unbalanced(chunks) Return (unbalanced_start, balanced, unbalanced_end), where each is
split_delete(chunks) Returns (stuff_before_DEL_START, stuff_inside_DEL_START_END,
locate_unbalanced_start(unbalanced_start,pre_delete,post_delete) pre_delete and post_delete implicitly point to a place in the
locate_unbalanced_end(unbalanced_end,pre_delete,post_delete) like locate_unbalanced_start, except handling end tags and
tokenize(html,include_hrefs=True) Parse the given HTML and returns token objects (words with attached tags).
parse_html(html,cleanup=True) Parses an HTML fragment, returning an lxml element. Note that the HTML will be
cleanup_html(html) This ‘cleans’ the HTML, meaning that any page structure is removed
split_trailing_whitespace(word) This function takes a word, such as ‘testnn’ and returns (‘test’,’nn’)
fixup_chunks(chunks) This function takes a list of chunks and produces a list of tokens.
flatten_el(el,include_hrefs,skip_tag=False) Takes an lxml element el, and generates all the text chunks for
split_words(text) Splits some text into words. Includes trailing whitespace
start_tag(el) The text representation of the start tag for a tag.
end_tag(el) The text representation of an end tag for a tag. Includes
is_word(tok)
is_end_tag(tok)
is_start_tag(tok)
fixup_ins_del_tags(html) Given an html string, move any <ins> or <del> tags inside of any
serialize_html_fragment(el,skip_outer=False) Serialize a single lxml element as HTML. The serialized form
_fixup_ins_del_tags(doc) fixup_ins_del_tags that works on an lxml document in-place
_contains_block_level_tag(el) True if the element contains any block-level elements, like <p>, <td>, etc.
_move_el_inside_block(el,tag) helper for _fixup_ins_del_tags; actually takes the <ins> etc tags
_merge_element_contents(el) Removes an element, but merges its contents into its place, e.g.,
default_markup(text, version)
html_annotate(doclist, markup=default_markup)

doclist should be ordered from oldest to newest, like:

>>> version1 = 'Hello World'
>>> version2 = 'Goodbye World'
>>> print(html_annotate([(version1, 'version 1'),
...                      (version2, 'version 2')]))
<span title="version 2">Goodbye</span> <span title="version 1">World</span>

The documents must be fragments (str/UTF8 or unicode), not complete documents

The markup argument is a function to markup the spans of words. This function is called like markup(‘Hello’, ‘version 2’), and returns HTML. The first argument is text and never includes any markup. The default uses a span with a title:

>>> print(default_markup('Some Text', 'by Joe'))
<span title="by Joe">Some Text</span>
tokenize_annotated(doc, annotation)

Tokenize a document and add an annotation attribute to each token

html_annotate_merge_annotations(tokens_old, tokens_new)

Merge the annotations from tokens_old into tokens_new, when the tokens in the new document already existed in the old document.

copy_annotations(src, dest)

Copy annotations from the tokens listed in src to the tokens in dest

compress_tokens(tokens)

Combine adjacent tokens when there is no HTML between the tokens, and they share an annotation

compress_merge_back(tokens, tok)

Merge tok into the last element of tokens (modifying the list of tokens in-place).

markup_serialize_tokens(tokens, markup_func)

Serialize the list of tokens into a list of text chunks, calling markup_func around text to add annotations.

htmldiff(old_html, new_html)

Do a diff of the old and new document. The documents are HTML fragments (str/UTF8 or unicode), they are not complete documents (i.e., no <html> tag).

Returns HTML with <ins> and <del> tags added around the appropriate text.

Markup is generally ignored, with the markup from new_html preserved, and possibly some markup from old_html (though it is considered acceptable to lose some of the old markup). Only the words in the HTML are diffed. The exception is <img> tags, which are treated like words, and the href attribute of <a> tags, which are noted inside the tag itself when there are changes.

htmldiff_tokens(html1_tokens, html2_tokens)

Does a diff on the tokens themselves, returning a list of text chunks (not tokens).

expand_tokens(tokens, equal=False)

Given a list of tokens, return a generator of the chunks of text for the data in the tokens.

merge_insert(ins_chunks, doc)

doc is the already-handled document (as a list of text chunks); here we add <ins>ins_chunks</ins> to the end of that.

class DEL_START
class DEL_END
class NoDeletes

Raised when the document no longer contains any pending deletes (DEL_START/DEL_END)

merge_delete(del_chunks, doc)

Adds the text chunks in del_chunks to the document doc (another list of text chunks) with marker to show it is a delete. cleanup_delete later resolves these markers into <del> tags.

cleanup_delete(chunks)

Cleans up any DEL_START/DEL_END markers in the document, replacing them with <del></del>. To do this while keeping the document valid, it may need to drop some tags (either start or end tags).

It may also move the del into adjacent tags to try to move it to a similar location where it was originally located (e.g., moving a delete into preceding <div> tag, if the del looks like (DEL_START, ‘Text</div>’, DEL_END)

split_unbalanced(chunks)

Return (unbalanced_start, balanced, unbalanced_end), where each is a list of text and tag chunks.

unbalanced_start is a list of all the tags that are opened, but not closed in this span. Similarly, unbalanced_end is a list of tags that are closed but were not opened. Extracting these might mean some reordering of the chunks.

split_delete(chunks)

Returns (stuff_before_DEL_START, stuff_inside_DEL_START_END, stuff_after_DEL_END). Returns the first case found (there may be more DEL_STARTs in stuff_after_DEL_END). Raises NoDeletes if there’s no DEL_START found.

locate_unbalanced_start(unbalanced_start, pre_delete, post_delete)

pre_delete and post_delete implicitly point to a place in the document (where the two were split). This moves that point (by popping items from one and pushing them onto the other). It moves the point to try to find a place where unbalanced_start applies.

As an example:

>>> unbalanced_start = ['<div>']
>>> doc = ['<p>', 'Text', '</p>', '<div>', 'More Text', '</div>']
>>> pre, post = doc[:3], doc[3:]
>>> pre, post
(['<p>', 'Text', '</p>'], ['<div>', 'More Text', '</div>'])
>>> locate_unbalanced_start(unbalanced_start, pre, post)
>>> pre, post
(['<p>', 'Text', '</p>', '<div>'], ['More Text', '</div>'])

As you can see, we moved the point so that the dangling <div> that we found will be effectively replaced by the div in the original document. If this doesn’t work out, we just throw away unbalanced_start without doing anything.

locate_unbalanced_end(unbalanced_end, pre_delete, post_delete)

like locate_unbalanced_start, except handling end tags and possibly moving the point earlier in the document.

class token

Represents a diffable token, generally a word that is displayed to the user. Opening tags are attached to this token when they are adjacent (pre_tags) and closing tags that follow the word (post_tags). Some exceptions occur when there are empty tags adjacent to a word, so there may be close tags in pre_tags, or open tags in post_tags.

We also keep track of whether the word was originally followed by whitespace, even though we do not want to treat the word as equivalent to a similar word that does not have a trailing space.

__new__(text, pre_tags=None, post_tags=None, trailing_whitespace="")
__repr__()
html()
class tag_token

Represents a token that is actually a tag. Currently this is just the <img> tag, which takes up visible space just like a word but is only represented in a document by a tag.

__new__(tag, data, html_repr, pre_tags=None, post_tags=None, trailing_whitespace="")
__repr__()
html()
class href_token

Represents the href in an anchor tag. Unlike other words, we only show the href when it changes.

html()
tokenize(html, include_hrefs=True)

Parse the given HTML and returns token objects (words with attached tags).

This parses only the content of a page; anything in the head is ignored, and the <head> and <body> elements are themselves optional. The content is then parsed by lxml, which ensures the validity of the resulting parsed document (though lxml may make incorrect guesses when the markup is particular bad).

<ins> and <del> tags are also eliminated from the document, as that gets confusing.

If include_hrefs is true, then the href attribute of <a> tags is included as a special kind of diffable token.

parse_html(html, cleanup=True)

Parses an HTML fragment, returning an lxml element. Note that the HTML will be wrapped in a <div> tag that was not in the original document.

If cleanup is true, make sure there’s no <head> or <body>, and get rid of any <ins> and <del> tags.

cleanup_html(html)

This ‘cleans’ the HTML, meaning that any page structure is removed (only the contents of <body> are used, if there is any <body). Also <ins> and <del> tags are removed.

split_trailing_whitespace(word)

This function takes a word, such as ‘testnn’ and returns (‘test’,’nn’)

fixup_chunks(chunks)

This function takes a list of chunks and produces a list of tokens.

flatten_el(el, include_hrefs, skip_tag=False)

Takes an lxml element el, and generates all the text chunks for that tag. Each start tag is a chunk, each word is a chunk, and each end tag is a chunk.

If skip_tag is true, then the outermost container tag is not returned (just its contents).

split_words(text)

Splits some text into words. Includes trailing whitespace on each word when appropriate.

start_tag(el)

The text representation of the start tag for a tag.

end_tag(el)

The text representation of an end tag for a tag. Includes trailing whitespace when appropriate.

is_word(tok)
is_end_tag(tok)
is_start_tag(tok)
fixup_ins_del_tags(html)

Given an html string, move any <ins> or <del> tags inside of any block-level elements, e.g. transform <ins><p>word</p></ins> to <p><ins>word</ins></p>

serialize_html_fragment(el, skip_outer=False)

Serialize a single lxml element as HTML. The serialized form includes the elements tail.

If skip_outer is true, then don’t serialize the outermost tag

_fixup_ins_del_tags(doc)

fixup_ins_del_tags that works on an lxml document in-place

_contains_block_level_tag(el)

True if the element contains any block-level elements, like <p>, <td>, etc.

_move_el_inside_block(el, tag)

helper for _fixup_ins_del_tags; actually takes the <ins> etc tags and moves them inside any block-level tags.

_merge_element_contents(el)

Removes an element, but merges its contents into its place, e.g., given <p>Hi <i>there!</i></p>, if you remove the <i> element you get <p>Hi there!</p>

class InsensitiveSequenceMatcher

Acts like SequenceMatcher, but tries not to find very small equal blocks amidst large spans of changes

get_matching_blocks()