needs.json builder performance

[danwos]

Some actions are happening also in the needs.json-builder, which are not needed for the json-output itself.

For instance in the function process_need_nodes the print_need_nodes()-func gets executed which prepares the docutil-table representation of a need.

sphinx-needs/sphinx_needs/directives/need.py

Line 397 in 01f0e5d

print_need_nodes(app, doctree, fromdocname, found_needs_nodes)

There may be other locations and actions, which calculate some unneeded stuff. Some ideas:

• Representation of needtable
• Representation of needflow and needuml
• Copying css/js data to _static (may be already deactivated)

We need a way to skip such actions if a builder is used, which is not transforming the doctree into a result-format.

A simple if builder == "needs" is not working, as it does not allow to integrated other builders in future.
For instance, a PR is getting prepared, which creates a needs.json per file or even per need-object.
So checking the builder-type may be a better approach, not sure....

[danwos]

Also smartquotes handling should be deactivated by default, as our jsin-files only contain the unrendered rst-code.