Add 6000 words of documentation on The Algorithm #298
Conversation
|
(Haven't actually tried building this, sleep time) |
book/src/algorithm.md
Outdated
| All of our analysis derives from the output of [cargo metadata][] and [cargo_metadata][]'s | ||
| interpretation of that, so it's worth discussing how we use it, and what we believe to | ||
| be true of its output. |
There was a problem hiding this comment.
I think this should probably be "our interpretation of that" rather than "cargo_metadata's interpretation of that"
book/src/algorithm.md
Outdated
|
|
||
| Cargo metadata produces the build graph in a kind of awkward way where some information | ||
| for the packages is in `"packages"` and some information is in `"resolve"`, and we need | ||
| to manually compute lots of facts like "roots", "only for tests", and "[topological sort][]" |
There was a problem hiding this comment.
the topological sort link here appears to be broken in github's renderer, but it might be fine in ours.
book/src/algorithm.md
Outdated
| and therefore require a desugarring to produce "fake" nodes, are *workspace members*. | ||
| These are the packages that will be tested if you run `cargo test --workspace`. | ||
|
|
||
| Actually doing this desugarring is really messy, because a lot of things about the "real" |
There was a problem hiding this comment.
(I could easily be wrong on this, but) Isn't it "desugaring" rather than "desugarring"?
| describe happens in a for loop). | ||
|
|
||
| If all we care about is finding out if a package has some criteria, then all | ||
| we need to do is run depth-first-search ([DFS][]) from the Root Node and see if it reaches |
There was a problem hiding this comment.
I'm not sure if the code is technically a depth-first search anymore given that it's now using a priority queue rather than a stack (since #296). Heaps don't preserve node order in the same way that a stack would.
I suppose it's more of a Best First Search but with the edge weights simplified because no successor can have better hereustics than its parent, and the hereustics have nothing to do with distance to target (though that is a potential thing we could do, by adding a distance into the sorting key)
book/src/algorithm.md
Outdated
| It's also worth noting that, due to audit-as-crates-io (and maybe other shenanigans | ||
| like patches?) you can end up in a situation where a third-party depends on a | ||
| first-party, or third-parties have `policy.criteria` entries. This is why | ||
| Step 3a is interleaved with Step 2. |
There was a problem hiding this comment.
Yup, [patch] can also do this.
|
The test failure is existing, and I've put up a fix in #301 |
No description provided.