The document's load_required method will reload if the parent_context's mtime is update. In the case of the NMR book, the root document is removed by two levels from the chapters. The chapter is always reloaded because the parent_mtime is newer because it's updated mtime is not set correctly.

src
├── fundamental_solnNMR
│   ├── fundamental_soln_nmr.dm
│   └── inept
│       └── inept.dm
├── index.dm

Implement a git interface in the CLI to work with the dissemia git servers

The list and outline environments should be able to process nested lists.

ex:

@outline{
- My first point
  - My second point
  - My third point
- My fourth point
}

Make a media manager that keeps track of included files and copies them to a media directory in the target folder.

Add block diagrams for processing flows in disseminate

Filenames that include special characters, like periods, trip up the \includegraphics latex function. This is fixed by wrapping the filename with curly braces.

The @code tag should support disseminate code

In html, block equations are rendered as inline equations. As a result, fractions are compressed to 1 line instead of multiple lines.

Dependency files that are not in the media/ directory of the project src are not copied correctly to the target paths.

current behavior

book/src/Chapter1/figures/fig1.png -> book/tex/fig1.png

expected behavior

book/src/Chapter1/figures/fig1.png -> book/tex/Chapter1/figures/fig1.png

Refactor document.documents_dict to use doc_ids instead of the src_filepath, as this will speed up some functions like the @ref tag

Implement build system inspired by Scons.

• Deciders based on modification times and md5 hashes
• A build environment that manages the builds for a project
• The build environment owns the root document, and the context stores a weakref to the build environment
• The environment manages builders for creating dependency files and target files
• The TargetBuilders render documents to their target format.
• CompositeBuilders can implement either ParallelBuilders for concurrent builds and SequentialBuilders for builds in sequence.

For html targets, it would be useful to have @next and @prev tags, which could be loaded in the header, to provide links to the previous and next documents using the document label.

Switch path management to pathlib.

Implement CLI boilerplate for projects and the option to change the template kind

Incorporate e-book formats, like epub, as a build target.

Implement a @cite tag that creates citation labels and has the option to lookup and format citation.

Add a validation function, is_valid, to tags to check that the contents are correct. This is a useful opportunity to raise a useful error message.

The server only reloads documents on the root tree page. The server should reload documents when project files are viewed.

Some context entries from the default (root) context, like @toc, are converted to tags for their respective documents. When sub-documents copy the parent context's entries, these tags should not be copied but rather reproduced from their original string entries. This will ensure that the converted tag will pertain and belong to each document's context, rather than the parent context.

Implement labels/refs for sidenotes and citations.

add functionality to create a glossary/index at the end of the document

The option for an output directory was removed from the CLI because it needs to handle multiple projects at once. For example, the CLI may be run for a root "books" directory with "book1" and "book2". Setting the output directory currently write the output of "book1" and "book2" to the same directory. An output option would need to write these separately.

When dependencies, like .asy files, are updated, those items should be re-rendered.

• Cleanup documentation examples to follow Table/Data
• Add a pygment parser for disseminate code

Implement signals to register and reset the label_manager and dependency_manager to decouple these from the document.

When compiling a textbook on linux, the converion from tex to pdf raises the following error on linux:

File "/usr/local/lib/python3.6/dist-packages/disseminate/convert/tex.py", line 60, in convert
    os.link(temp_filepath_pdf, self.target_filepath())
OSError: [Errno 18] Invalid cross-device link: '/tmp/tmp0k6awsdv/index.pdf' -> '/builds/psnee/free-energy/public/pdf/Chapter1_GasEquationsState/index.pdf'

Fix benchmark regression on commit 0acc136.

Documents with a toc entry in the header and a toc entry in the body appear to have their toc-levels misset. i.e. the header toc entry starts at toc-level-1 and the toc in the body of the document starts at the last toc-level from the header toc. However, the body toc should start at toc-level-1 too.

The current dependency manager operates with locking of each file, meaning that it can't robustly run the conversion of files concurrently.

Solutions:

• Implement concurrently in the conversions
• Use an external tool
  • scons seems like a natural choice, but it doesn't integrate into python easily. It's intended to run SCons files.

Context entries are sometimes copied, not copied, appended or replaced. Some of these options are currently available, like do not inherit, but it's not clear whether inherited attributes are replaced (immutables) or appended (mutables). Options in the context, like do_not_inherit, should make this more clear.

An option to fix this would be something like a replace attribute that lists entries that are replaced when present in the child context, rather than appended.

Implement a @panel tag for subfigures.

Equation text converted to pdf/svg (\text) produces grainy text.

Implement lxml namespaces so that attributes like epub:type can be included.

This could be implemented as attributes with dicts as values.

Add a check for the parent context's mtime to force a reload for sub-documents.

Implemented in the feature/table branch

The @toc tag in a root document is not updated when a new subdocument is added.

Exceptions should include the line number from the source file.

The navigation tags (@prev, @next) can be sped up if these are calculated as a processor for the context when creating documents. The document list for each target would only need to be loaded once to populate these tags.

However, this would likely need to be done after all documents are loaded as a "post-load" signal, whereas the other processors are run as "on-load" signals.

The @prev/@next tags don't show up for some of the chapters in Anna Karenina. For example, Chap 34 of Part One doesn't show a @next tag.

It may be useful to have an auto_image_width option in the context to set the width of images if a width is not specified.

It is convenient to have values stored in the header, like the document's title or author names, and substitute these into the main document. The text of these can be substituted like a macro, or a Substitution tag can be used to refer back to the header/context entry. Each of these approaches have specific strengths:

Substitution

Advantages

• For a substitution tag, an entry may be a tag itself. Having the entry as a tag is useful when using templates, as these can be used and formatted in the template, separately from the body of the document.

(( authors.html ))
(( body.html ))

Disadvantages

• Substitutions may not work for header/context entries with the same name as tags. For example, 'title' can be used as a Branch tag, but it cannot be used as a substitution later in the document, as the 'title' tag will be used instead of a substitution tag.
• It's fairly easy to create circular relationships. Mechanisms that block circular relationships may not properly substitute the contents of the Substitution tag

Macros

Advantages

• Much simpler to work with
• A macro could be processed into an ast after the macro substitution.

(( authors.html ))
(( body.html ))

## Disadvantages
+ Line numbering relating to the source document is difficult to track