-
Notifications
You must be signed in to change notification settings - Fork 6
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
feat: Split methods between HugrMut
and HugrMutInternals
#441
Conversation
- split methods between HugrMutInternals & HugrMut - Add a hierarchy root to `Hugr::canonical_order` - check the root type when querying a hugrview - Add a lifetime to `HugrInternals::Portgraph` so the views can return clones instead of double references.
It clashes with the new convexCheck in portgraph since it requires `Copy` on the LinkView :/
co-authored-by: Seyon Sivarajah <seyon.sivarajah@quantinuum.com>
I'll re-add the rolled-back lifetime parameters in |
fn root(&self) -> Node; | ||
/// Return the root node of this view. | ||
#[inline] | ||
fn root(&self) -> Node { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is just a way of exposing the hidden-in-HugrInternals root_node, right?
(Might one call that root
too and then do (self as HugrInternals).root()
here?)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yeah, I don't seriously suggest that, think I've got it now
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The different name is to avoid ambiguity at local call sites. Users shouldn't see the HugrMutInternals
, so there shouldn't be any confusion.
@@ -340,6 +384,12 @@ pub enum HugrError { | |||
/// An error occurred while manipulating the hierarchy. | |||
#[error("An error occurred while manipulating the hierarchy.")] | |||
HierarchyError(#[from] portgraph::hierarchy::AttachError), | |||
/// The node doesn't exist. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I wonder about the utility of naming the single field within, but not too worried
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I prefer to use named fields when the order is not relevant (what does x.0
mean?). But I have no strong opinion on this.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If there are multiple fields of the same type (e.g. expected
and actual
) then very much agree that named fields is the way to go. But here I don't think the name adds anything beyond what the type says already, and I really can't think of much else (of any type) you could put in an InvalidNode... (a validity-reason, perhaps. So ok if you prefer it this way!)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I changed it into a tuple struct for consistency with the other errors.
I'd also mention in the description - remove |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Nice!! :)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Nice! Like it, and think it does what I'm after, thanks @aborgna-q :-)
} | ||
|
||
/// Add a node to the graph. | ||
fn add_node(&mut self, nodetype: NodeType) -> Node { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is the only implementation of add_node
, I wonder if it really belongs here or might be better as a method on Hugr
itself?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I thought for a moment that the same might be true for all of these HugrMutInternals
methods, but no - add_node
and add_op
are the only ones that don't do (or aren't part of) a valid_xxx
check, so I think that's a reasonable argument that they belong somewhere else.
*self.get_metadata_mut(node) = metadata; | ||
/// Returns a [`HugrError::InvalidNode`] otherwise. | ||
#[inline] | ||
fn valid_node(&self, node: Node) -> Result<(), HugrError> { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Agreed with the implementation; the naming here is rather tricky. I don't really have a better suggestion (I thought about valid_parent
instead of valid_node
and valid_sibling
instead of valid_descendant
, and I think this is better in some places - but worse in others, so not really any better overall). So ok!
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hmmm, how about valid_non_root
rather than valid_descendant
? These mean pretty much the same thing, but descendant
might be confusing in a flat view that has only children and nothing deeper? (And valid_non_root
also describes quite well what the implementation actually checks...)
@@ -340,6 +384,12 @@ pub enum HugrError { | |||
/// An error occurred while manipulating the hierarchy. | |||
#[error("An error occurred while manipulating the hierarchy.")] | |||
HierarchyError(#[from] portgraph::hierarchy::AttachError), | |||
/// The node doesn't exist. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If there are multiple fields of the same type (e.g. expected
and actual
) then very much agree that named fields is the way to go. But here I don't think the name adds anything beyond what the type says already, and I really can't think of much else (of any type) you could put in an InvalidNode... (a validity-reason, perhaps. So ok if you prefer it this way!)
## 🤖 New release * `quantinuum-hugr`: 0.1.0 <details><summary><i><b>Changelog</b></i></summary><p> <blockquote> ## 0.1.0 (2024-01-15) ### Bug Fixes - Subgraph boundaries with copies ([#440](#440)) - [**breaking**] Use internal tag for SumType enum serialisation ([#462](#462)) - Check kind before unwrap in insert_identity ([#475](#475)) - Allow for variables to get solved in inference ([#478](#478)) - In IdentityInsertion add noop to correct parent ([#477](#477)) - Failing release tests ([#485](#485)) - [**breaking**] Serialise `Value`, `PrimValue`, and `TypeArg` with internal tags ([#496](#496)) - Serialise custom constants with internal tag ([#502](#502)) - [**breaking**] Reduce max int width in arithmetic extension to 64 ([#504](#504)) - HugrView::get_function_type ([#507](#507)) - TODO in resolve_extension_ops: copy across input_extensions ([#510](#510)) - Use given input extensions in `define_function` ([#524](#524)) - Lessen requirements for hugrs in outline_cfg ([#528](#528)) - Make unification logic less strict ([#538](#538)) - Simple replace incorrectly copying metadata ([#545](#545)) - Account for self-referencial constraints ([#551](#551)) - Consider shunted metas when comparing equality ([#555](#555)) - Join labels in issue workflow ([#563](#563)) - Comment out broken priority code ([#562](#562)) - Handling of issues with no priority label ([#573](#573)) - Don't insert temporary wires when extracting a subgraph ([#582](#582)) - Improve convexity checking and fix test ([#585](#585)) - Ignore input->output links in SiblingSubgraph::try_new_dataflow_subgraph ([#589](#589)) - Enforce covariance of SiblingMut::RootHandle ([#594](#594)) - Erratic stack overflow in infer.rs (live_var) ([#638](#638)) - Work harder in variable instantiation ([#591](#591)) - Actually add the error type to prelude ([#672](#672)) - Serialise dynamically computed opaqueOp signatures ([#690](#690)) - FuncDefns don't require that their extensions match their children ([#688](#688)) - Binary compute_signature returning a PolyFuncType with binders ([#710](#710)) - Use correct number of args for int ops ([#723](#723)) - [**breaking**] Add serde tag to TypeParam enum ([#722](#722)) - Allow widening and narrowing to same width. ([#735](#735)) - Case node should not have an external signature ([#749](#749)) - [**breaking**] Normalize input/output value/static/other ports in `OpType` ([#783](#783)) - No dataflow_signature for block types ([#792](#792)) - Ignore unsupported test in miri ([#794](#794)) - Include schema rather than read file ([#807](#807)) ### Documentation - Add operation constraint to quantum extension ([#543](#543)) - Coverage check section in DEVELOPMENT.md ([#648](#648)) - Remove "quantum extension" from HUGR spec. ([#694](#694)) - Improve crate-level docs, including example code. ([#698](#698)) - Spec cleanups and clarifications ([#742](#742)) - Spec clarifications ([#738](#738)) - Spec updates ([#741](#741)) - [spec] Remove references to causal cone and Order edges from Input ([#762](#762)) - Mention experimental inference in readme ([#800](#800)) - Collection of spec updates for 0.1 ([#801](#801)) - Add schema v0 ([#805](#805)) - Update spec wrt. polymorphism ([#791](#791)) ### Features - Derive things for builder structs ([#229](#229)) - Return a slice of types from the signature ([#238](#238)) - Move `dot_string` to `HugrView` ([#271](#271)) - [**breaking**] Change `TypeParam::USize` to `TypeParam::BoundedNat` and use in int extensions ([#445](#445)) - TKET2 compatibility requirements ([#450](#450)) - Split methods between `HugrMut` and `HugrMutInternals` ([#441](#441)) - Add `HugrView::node_connections` to get all links between nodes ([#460](#460)) - Location information in extension inference error ([#464](#464)) - Add T, Tdg, X gates ([#466](#466)) - [**breaking**] Add `ApplyResult` associated type to `Rewrite` ([#472](#472)) - Implement rewrite `IdentityInsertion` ([#474](#474)) - Instantiate inferred extensions ([#461](#461)) - Default DFG builders to open extension sets ([#473](#473)) - Some helper methods ([#482](#482)) - Extension inference for conditional nodes ([#465](#465)) - Add ConvexChecker ([#487](#487)) - Add clippy lint for mut calls in a debug_assert ([#488](#488)) - Default more builder methods to open extension sets ([#492](#492)) - Port is serializable ([#489](#489)) - More general portgraph references ([#494](#494)) - Add extension deltas to CFG ops ([#503](#503)) - Implement petgraph traits on Hugr ([#498](#498)) - Make extension delta a parameter of CFG builders ([#514](#514)) - [**breaking**] SiblingSubgraph does not borrow Hugr ([#515](#515)) - Validate TypeArgs to ExtensionOp ([#509](#509)) - Derive Debug & Clone for `ExtensionRegistry`. ([#530](#530)) - Const nodes are built with some extension requirements ([#527](#527)) - Some python errors and bindings ([#533](#533)) - Insert_hugr/insert_view return node map ([#535](#535)) - Add `SiblingSubgraph::try_from_nodes_with_checker` ([#547](#547)) - PortIndex trait for undirected port parameters ([#553](#553)) - Insert/extract subgraphs from a HugrView ([#552](#552)) - Add `new_array` operation to prelude ([#544](#544)) - Add a `DataflowParentID` node handle ([#559](#559)) - Make TypeEnum and it's contents public ([#558](#558)) - Optional direction check when querying a port index ([#566](#566)) - Extension inference for CFGs ([#529](#529)) - Better subgraph verification errors ([#587](#587)) - Compute affected nodes for `SimpleReplacement` ([#600](#600)) - Move `SimpleReplace::invalidation_set` to the `Rewrite` trait ([#602](#602)) - [**breaking**] Resolve extension ops (mutating Hugr) in (infer_and_->)update_validate ([#603](#603)) - Add accessors for node index and const values ([#605](#605)) - [**breaking**] Expose the value of ConstUsize ([#621](#621)) - Nicer debug and display for core types ([#628](#628)) - [**breaking**] Static checking of Port direction ([#614](#614)) - Builder and HugrMut add_op_xxx default to open extensions ([#622](#622)) - [**breaking**] Add panicking integer division ops ([#625](#625)) - Add hashable `Angle` type to Quantum extension ([#608](#608)) - [**breaking**] Remove "rotations" extension. ([#645](#645)) - Port.as_directed to match on either direction ([#647](#647)) - Impl GraphRef for PetgraphWrapper ([#651](#651)) - Provide+implement Replace API ([#613](#613)) - Require the node's metadata to always be a Map ([#661](#661)) - [**breaking**] Polymorphic function types (inc OpDefs) using dyn trait ([#630](#630)) - Make prelude error type public ([#669](#669)) - Shorthand for retrieving custom constants from `Const`, `Value` ([#679](#679)) - [**breaking**] HugrView API improvements ([#680](#680)) - Make FuncDecl/FuncDefn polymorphic ([#692](#692)) - [**breaking**] Simplify `SignatureFunc` and add custom arg validation. ([#706](#706)) - [**breaking**] Drop the `pyo3` feature ([#717](#717)) - [**breaking**] `OpEnum` trait for common opdef functionality ([#721](#721)) - MakeRegisteredOp trait for easier registration ([#726](#726)) - Getter for `PolyFuncType::body` ([#727](#727)) - `Into<OpType>` for custom ops ([#731](#731)) - Always require a signature in `OpaqueOp` ([#732](#732)) - Values (and hence Consts) know their extensions ([#733](#733)) - [**breaking**] Use ConvexChecker trait ([#740](#740)) - Custom const for ERROR_TYPE ([#756](#756)) - Implement RemoveConst and RemoveConstIgnore ([#757](#757)) - Constant folding implemented for core and float extension ([#769](#769)) - Constant folding for arithmetic conversion operations ([#720](#720)) - DataflowParent trait for getting inner signatures ([#782](#782)) - Constant folding for logic extension ([#793](#793)) - Constant folding for list operations ([#795](#795)) - Add panic op to prelude ([#802](#802)) - Const::from_bool function ([#803](#803)) ### HugrMut - Validate nodes for set_metadata/get_metadata_mut, too ([#531](#531)) ### HugrView - Validate nodes, and remove Base ([#523](#523)) ### Miscellaneous Tasks - [**breaking**] Update portgraph 0.10 and pyo3 0.20 ([#612](#612)) - [**breaking**] Hike MSRV to 1.75 ([#761](#761)) ### Performance - Use lazy static definittion for prelude registry ([#481](#481)) ### Refactor - Move `rewrite` inside `hugr`, `Rewrite` -> `Replace` implementing new 'Rewrite' trait ([#119](#119)) - Use an excluded upper bound instead of max log width. ([#451](#451)) - Add extension info to `Conditional` and `Case` ([#463](#463)) - `ExtensionSolution` only consists of input extensions ([#480](#480)) - Remove builder from more DFG tests ([#490](#490)) - Separate hierarchy views ([#500](#500)) - [**breaking**] Use named struct for float constants ([#505](#505)) - Allow NodeType::new to take any Into<Option<ExtensionSet>> ([#511](#511)) - Move apply_rewrite from Hugr to HugrMut ([#519](#519)) - Use SiblingSubgraph in SimpleReplacement ([#517](#517)) - CFG takes a FunctionType ([#532](#532)) - Remove check_custom_impl by inlining into check_custom ([#604](#604)) - Insert_subgraph just return HashMap, make InsertionResult new_root compulsory ([#609](#609)) - [**breaking**] Rename predicate to TupleSum/UnitSum ([#557](#557)) - Simplify infer.rs/report_mismatch using early return ([#615](#615)) - Move the core types to their own module ([#627](#627)) - Change &NodeType->&OpType conversion into op() accessor ([#623](#623)) - Infer.rs 'fn results' ([#631](#631)) - Be safe ([#637](#637)) - NodeType constructors, adding new_auto ([#635](#635)) - Constraint::Plus stores an ExtensionSet, which is a BTreeSet ([#636](#636)) - [**breaking**] Remove `SignatureDescription` ([#644](#644)) - [**breaking**] Remove add_op_<posn> by generalizing add_node_<posn> with "impl Into" ([#642](#642)) - Rename accidentally-changed Extension::add_node_xxx back to add_op ([#659](#659)) - [**breaking**] Remove quantum extension ([#670](#670)) - Use type schemes in extension definitions wherever possible ([#678](#678)) - [**breaking**] Flatten `Prim(Type/Value)` in to parent enum ([#685](#685)) - [**breaking**] Rename `new_linear()` to `new_endo()`. ([#697](#697)) - Replace NodeType::signature() with io_extensions() ([#700](#700)) - Validate ExtensionRegistry when built, not as we build it ([#701](#701)) - [**breaking**] One way to add_op to extension ([#704](#704)) - Remove Signature struct ([#714](#714)) - Use `MakeOpDef` for int_ops ([#724](#724)) - [**breaking**] Use enum op traits for floats + conversions ([#755](#755)) - Avoid dynamic dispatch for non-folding operations ([#770](#770)) - Simplify removeconstignore verify ([#768](#768)) - [**breaking**] Unwrap BasicBlock enum ([#781](#781)) - Make clear const folding only for leaf ops ([#785](#785)) - [**breaking**] `s/RemoveConstIgnore/RemoveLoadConstant` ([#789](#789)) - Put extension inference behind a feature gate ([#786](#786)) ### SerSimpleType - Use Vec not TypeRow ([#381](#381)) ### SimpleReplace+OutlineCfg - Use HugrMut methods rather than .hierarchy/.graph ([#280](#280)) ### Testing - Update inference test to not use DFG builder ([#550](#550)) - Strengthen "failing_sccs_test", rename to "sccs" as it's not failing! ([#660](#660)) - [**breaking**] Improve coverage in signature and validate ([#643](#643)) - Use insta snapshots to add dot_string coverage ([#682](#682)) - Miri ignore file-opening test ([#684](#684)) - Unify the serialisation tests ([#730](#730)) - Add schema validation to roundtrips ([#806](#806)) ### `ConstValue - :F64` and `OpaqueOp::new` ([#206](#206)) ### Cleanup - Remove outdated comment ([#536](#536)) ### Cosmetic - Format + remove stray TODO ([#444](#444)) ### Doc - Crate name as README title + add reexport ([#199](#199)) ### S/EdgeKind - :Const/EdgeKind::Static/ ([#201](#201)) ### Simple_replace.rs - Use HugrMut::remove_node, includes clearing op_types ([#242](#242)) ### Spec - Remove "Draft 3" from title of spec document. ([#590](#590)) - Rephrase confusing paragraph about TailLoop inputs/outputs ([#567](#567)) ### Src/ops/validate.rs - Common-up some type row calculations ([#254](#254)) </blockquote> </p></details> --- This PR was generated with [release-plz](https://github.com/MarcoIeni/release-plz/). --------- Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Agustín Borgna <121866228+aborgna-q@users.noreply.github.com>
Splits the mutable methods between
HugrMutInternals
&HugrMut
.When implemented by a (future) non-
AsMut<Hugr>
, the default impl checks whether the related nodes are valid in the view before calling the internal hugrs mutable primitives.Other small changes:
canonicalize_nodes
fromHugrMut
toHugr
, since it doesn't make sense to call in on a view.Hugr::canonical_order
so we can get the canonical order for regions.Add a lifetime toHugrInternals::Portgraph
so the views can return clones instead of double references.I had to roll it back due to the convex check requiring
Copy
on the portgraph view.