The code in these Mix projects was extracted from a much larger system, in order to have smaller and more cohesive versions for experimentation. In the original system, the code adds a table of contents (TOC) to the HTML output page from a Phoenix-based server.
Both versions do this task successfully, but the old version
causes an error for mix dialyzer (see dial_out for details).
The current version (see take_3, below) passes Dialyzer.
This code uses several functions from the Floki library:
find/2- search the input tree for atocelementparse_document/1- generate a parse tree from the input HTMLparse_fragment/1- generate a parse tree from the TOC HTMLraw_html/1- regenerate HTML from the parse treetext/1- extract text from header elementstraverse_and_update/2- replace thetocelement with the TOCtraverse_and_update/3- add the TOC to the page
It also uses some of Floki's type definitions, in its @spec entries:
Floki.html_tagFloki.html_tree
Note:
Floki's traverse_and_update/3 function is very new
and its @spec entry has a small error.
So, for the moment,
I'm using a modified copy of the library.
This Git repo contains two Mix projects.
Each of these passes mix test, but fails mix dialyzer.
(See dial_out_t1 and dial_out_t2 for details.)
The projects differ as discussed below.
The smalltoc.ex file was copied
from the old router.toc.ex in the larger system.
The Phoenix-related code was then removed
and a test file (smalltoc_test.exs) was created.
This version of Smalltoc.add/1:
- wraps each HTML header in a
<a name=...element - replaces the
<toc />element with a TOC list
The smalltoc.ex file was edited
to remove the TOC generation and replacement code;
the test file was modified accordingly.
The smalltoc.ex file was copied
from the new router.toc.ex in the larger system.
The Phoenix-related code was then removed
and a test file (smalltoc_test.exs) was created.
This version of Smalltoc.add/1 uses a combination of Floki calls
and regular expressions.
The resulting code is shorter, simpler, and passes Dialyzer.