This Engineering Notebook post announces several momentous Ahas, all
directly related to VNodes and our attitude towards them. These Ahas
resolve long-standing questions and suggest new ways of simplifying Leo's
most fundamental code.
*Acknowledgements*: These Ahas would not have happened with recent
discussions with Félix and Vitalije.
*Aha! *The best way to check outline structure is to test the contents of
the *v.parents* and *v.children* arrays.
The new *c.checkVnodeLinks* method is 100 times faster than previous checks
and finds problems that have lain hidden until now. Leo performs these
checks when reading, pasting or saving outlines.
The read check *has already failed* when loading an outline after switching
branches. The paste checks cause two paste-related unit tests to fail.
*Aha!* c.checkVnodeLinks can *repair* broken or missing links.
The unified PR contains prototype code. I'll enable this code only after it
passes a strict new unit test.
*Aha!* VNodes are the proper context for most low-level operations.
I've never realized how true this is until now.
*Aha!* The *c.all_unique_nodes* generator (or something similar, see below)
should be the basis for many VNode operations.
Relying on c.all_unique_nodes is a new pattern for me! For example,
c.checkVnodeLinks now uses c.all_unique_nodes instead of a bespoke
recursion on v.children.
*Aha!* I now know how to backup/restore a VNode and its descendants!
Neither recursion nor pickling will work because VNodes contain backlinks
in their v.parents arrays. Instead, Leo's new archive code will create a
json-like dict in which almost all keys and values will be gnxs!
Define *p.archive* (yes, p) something like this (untested, but the code
passes mypy):
def archive(self) -> dict[str, Any]:
"""Return a json-like archival dictionary for p/v.unarchive."""
p = self
# Create a list of all vnodes in p.self_and_subtree.
all_unique_vnodes: list[VNode] = []
for p in p.self_and_subtree():
if p.v not in all_unique_vnodes:
all_unique_vnodes.append(p.b)
# Create an archive of all_vnodes.
parents_dict: dict[str, list[str]] = {}
for v in all_unique_vnodes:
parents_dict [v.gnx] = [z.gnx for z in v.parents]
children_dict: dict[str, list[str]] = {}
for v in all_unique_vnodes:
children_dict [v.gnx] = [z.gnx for z in v.children]
marks_dict: dict[str, str] = {}
for v in all_unique_vnodes:
marks_dict [v.gnx] = str(int(v.isMarked()))
uas_dict: dict[str, dict] = {}
for v in all_unique_vnodes:
uas_dict [v.gnx] = v.archive_ua() # To do.
return {
'vnodes': all_unique_vnodes,
'parents': parents_dict,
'children': children_dict,
'marks': marks_dict,
'uAs': uas_dict,
}
*Summary*
*c.checkVnodeLinks* reveals previous hidden errors in Leo's code. Leo's
codebase is about to become significantly more solid!
c.checkVnodeLinks can correct any broken or missing links in a
straightforward way. c.checkVnodeLinks will raise ValueError during unit
tests :-)
A generator on unique vnodes is the natural way to accumulate archival
data. These data should be a json-like dict in which all keys will be gnxs
and many values will be lists of gnxs.
*p.archive* and *v.archive_uas* will likely replace ad-hoc code scattered
throughout Leo's codebase.
*I shall merge the unified PR into devel later today.* There is no reason
to delay. I'll then create new issues and PRs:
- Fix the failing unit tests related to pasting nodes.
- Create a stringent new unit test for repairing links.
- Enable automatic link repair.
- Experiment with archives.
Edward
--
You received this message because you are subscribed to the Google Groups
"leo-editor" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/leo-editor/a5a37019-1a2b-4012-9849-cc903e442451n%40googlegroups.com.
