While implementing caching for components, I did some cleanup. Despite
the fact that subgraph caching is put on hold for now (in favor of a
larger cache refactor later), these are the changes that I think are
worth keeping anyway.
* Makes subgraph node IDs deterministic
* Allows usage of the topological sort without execution
* Tracks parent nodes (i.e. those that caused a node to be created) and
display nodes (i.e. the one we want to highlight while an ephemeral
node is executing) separately.
Execution blocking can be done by returning an `ExecutionBlocker`
(imported from graph_utils) either in place of results or as a specific
output. Any node that uses an `ExecutionBlocker` as input will be
skipped. This operates on a per-entry basis when inputs are lists.
If the `ExecutionBlocker` is initialized with an error message, that
message will be displayed on the first node it's used on (and further
downstream nodes will be silently skipped).
Honestly, I'm still a little concerned here. There's nothing stopping a
custom node from having a data type of ["str",int]. I've improved
recognition to at least prevent the detection of other types, but we
may still want a more systemic fix (e.g. wrapping literals within a
class when using them as inputs to nodes in subgraphs).
This PR inverts the execution model -- from recursively calling nodes to
using a topological sort of the nodes. This change allows for
modification of the node graph during execution. This allows for two
major advantages:
1. The implementation of lazy evaluation in nodes. For example, if a
"Mix Images" node has a mix factor of exactly 0.0, the second image
input doesn't even need to be evaluated (and visa-versa if the mix
factor is 1.0).
2. Dynamic expansion of nodes. This allows for the creation of dynamic
"node groups". Specifically, custom nodes can return subgraphs that
replace the original node in the graph. This is an *incredibly*
powerful concept. Using this functionality, it was easy to
implement:
a. Components (a.k.a. node groups)
b. Flow control (i.e. while loops) via tail recursion
c. All-in-one nodes that replicate the WebUI functionality
d. and more
All of those were able to be implemented entirely via custom nodes
without hooking or replacing any core functionality. Within this PR,
I've included all of these proof-of-concepts within a custom node pack.
In reality, I would expect some number of them to be merged into the
core node set (with the rest left to be implemented by custom nodes).
I made very few changes to the front-end, so there are probably some
easy UX wins for someone who is more willing to wade into .js land. The
user experience is a lot better than I expected though -- progress shows
correctly in the UI over the nodes that are being expanded.
