Procházet zdrojové kódy

Revise/add to new limitations section.

tags/v0.5
Ben Kurtovic před 7 roky
rodič
revize
784e5e7b8d
4 změnil soubory, kde provedl 78 přidání a 38 odebrání
  1. +32
    -20
      README.rst
  2. +0
    -17
      docs/caveats.rst
  3. +1
    -1
      docs/index.rst
  4. +45
    -0
      docs/limitations.rst

+ 32
- 20
README.rst Zobrazit soubor

@@ -115,36 +115,47 @@ Likewise, use ``unicode(code)`` in Python 2.

Limitations
-----------
While the MediaWiki parser generates HTML, mwparserfromhell acts as an interface to
the source code. mwparserfromhell therefore is unaware of template definitions since
if it would substitute templates with their output you would no longer be working
with the source code. This has several implications:

* Start and end tags generated by templates aren't recognized e.g. ``<b>foobar{{bold-end}}``.
While the MediaWiki parser generates HTML and has access to the contents of
templates, among other things, mwparserfromhell acts as a direct interface to
the source code only. This has several implications:

* Templates adjacent to external links e.g. ``http://example.com{{foo}}`` are
considered part of the link.
* Syntax elements produced by a template transclusion cannot be detected. For
example, imagine a hypothetical page ``"Template:End-bold"`` that contained
the text ``</b>``. While MediaWiki would correctly understand that
``<b>foobar{{end-bold}}`` translates to ``<b>foobar</b>``, mwparserfromhell
has no way of examining the contents of ``{{end-bold}}``. Instead, it would
treat the bold tag as unfinished, possibly extending further down the page.

* Crossed constructs like ``{{echo|''Hello}}, world!''`` are not supported,
the first node is treated as plain text.
* Templates adjacent to external links, as in ``http://example.com{{foo}}``,
are considered part of the link. In reality, this would depend on the
contents of the template.

The current workaround for cases where you are not interested in text
formatting is to pass ``skip_style_tags=True`` to ``mwparserfromhell.parse()``.
This treats ``''`` and ``'''`` like plain text.
* When different syntax elements cross over each other, as in
``{{echo|''Hello}}, world!''``, the parser gets confused because this cannot
be represented by an ordinary syntax tree. Instead, the parser will treat the
first syntax construct as plain text. In this case, only the italic tag would
be properly parsed.

A future version of mwparserfromhell will include multiple parsing modes to get
around this restriction.
**Workaround:** Since this commonly occurs with text formatting and text
formatting is often not of interest to users, you may pass
*skip_style_tags=True* to ``mwparserfromhell.parse()``. This treats ``''``
and ``'''`` as plain text.

Configuration unawareness
A future version of mwparserfromhell may include multiple parsing modes to
get around this restriction more sensibly.

* `word-ending links`_ are not supported since the linktrail rules are language-specific.
Additionally, the parser lacks awareness of certain wiki-specific settings:

* Localized namespace names aren't recognized, e.g. ``[[File:...]]``
links are treated as regular wikilinks.
* `word-ending links`_ are not supported, since the linktrail rules are
language-specific.

* Anything that looks like an XML tag is parsed as a tag
since the available tags are extension-dependent.
* Localized namespace names aren't recognized, so file links (such as
``[[File:...]]``) are treated as regular wikilinks.

* Anything that looks like an XML tag is treated as a tag, even if it is not a
recognized tag name, since the list of valid tags depends on loaded MediaWiki
extensions.

Integration
-----------


+ 0
- 17
docs/caveats.rst Zobrazit soubor

@@ -1,17 +0,0 @@
Caveats
=======

An inherent limitation in wikicode prevents us from generating complete parse
trees in certain cases. For example, the string ``{{echo|''Hello}}, world!''``
produces the valid output ``<i>Hello, world!</i>`` in MediaWiki, assuming
``{{echo}}`` is a template that returns its first parameter. But since
representing this in mwparserfromhell's node tree would be impossible, we
compromise by treating the first node (i.e., the template) as plain text,
parsing only the italics.

The current workaround for cases where you are not interested in text
formatting is to pass *skip_style_tags=True* to :func:`mwparserfromhell.parse`.
This treats ``''`` and ``'''`` like plain text.

A future version of mwparserfromhell will include multiple parsing modes to get
around this restriction.

+ 1
- 1
docs/index.rst Zobrazit soubor

@@ -40,7 +40,7 @@ Contents
:maxdepth: 2

usage
caveats
limitations
integration
changelog
API Reference <api/modules>


+ 45
- 0
docs/limitations.rst Zobrazit soubor

@@ -0,0 +1,45 @@
Limitations
===========

While the MediaWiki parser generates HTML and has access to the contents of
templates, among other things, mwparserfromhell acts as a direct interface to
the source code only. This has several implications:

* Syntax elements produced by a template transclusion cannot be detected. For
example, imagine a hypothetical page ``"Template:End-bold"`` that contained
the text ``</b>``. While MediaWiki would correctly understand that
``<b>foobar{{end-bold}}`` translates to ``<b>foobar</b>``, mwparserfromhell
has no way of examining the contents of ``{{end-bold}}``. Instead, it would
treat the bold tag as unfinished, possibly extending further down the page.

* Templates adjacent to external links, as in ``http://example.com{{foo}}``,
are considered part of the link. In reality, this would depend on the
contents of the template.

* When different syntax elements cross over each other, as in
``{{echo|''Hello}}, world!''``, the parser gets confused because this cannot
be represented by an ordinary syntax tree. Instead, the parser will treat the
first syntax construct as plain text. In this case, only the italic tag would
be properly parsed.

**Workaround:** Since this commonly occurs with text formatting and text
formatting is often not of interest to users, you may pass
*skip_style_tags=True* to ``mwparserfromhell.parse()``. This treats ``''``
and ``'''`` as plain text.

A future version of mwparserfromhell may include multiple parsing modes to
get around this restriction more sensibly.

Additionally, the parser lacks awareness of certain wiki-specific settings:

* `word-ending links`_ are not supported, since the linktrail rules are
language-specific.

* Localized namespace names aren't recognized, so file links (such as
``[[File:...]]``) are treated as regular wikilinks.

* Anything that looks like an XML tag is treated as a tag, even if it is not a
recognized tag name, since the list of valid tags depends on loaded MediaWiki
extensions.

.. _word-ending links: https://www.mediawiki.org/wiki/Help:Links#linktrail

Načítá se…
Zrušit
Uložit