feat: Upgrade Hugr and start using the shared Pydantic model #201

mark-koch · 2024-05-02T13:47:45Z

This PR makes Guppy depend on the Hugr 0.4 release candidate 0.4.0-alpha.1. Closes #198

Most of the edits are related to changes in the Pydantic model. However, we also had to get rid of TypeApply since this no longer exists in Hugr. Therefore, Guppy now rejects all programs that use polymorphic functions as values if the type arguments cannot be inferred

mark-koch · 2024-05-02T13:59:45Z

guppylang/compiler/cfg_compiler.py

-                    [v for v in row if not v.ty.linear or is_return_var(v.name)]
+                    [
+                        v
+                        for v in sort_vars(row)
+                        if not v.ty.linear or is_return_var(v.name)
+                    ]


I moved the call to sort_vars from choose_vars_for_tuple_sum() below to here as a drive-by since this is more consistent with the other sorting logic in this method

mark-koch · 2024-05-02T14:01:22Z

guppylang/compiler/expr_compiler.py

-            if any(value is None for value in vs):
-                return None
-            return val.Tuple(vs=vs)
+            if doesnt_contain_none(vs):
+                return ops.Value(ops.TupleValue(vs=vs))


Mypy started complaining about this, so I added a doesnt_contain_none type guard

mark-koch · 2024-05-02T14:03:08Z

guppylang/definition/custom.py

-        if self.ty.parametrized:
-            raise InternalGuppyError(
-                "Can't yet generate higher-order versions of custom functions. This "
-                "requires generic function *definitions*"
-            )


A nice benefit of having the type args available here is that we can properly handle this case now!

mark-koch · 2024-05-02T14:06:17Z

tests/integration/test_poly.py

@@ -259,6 +261,7 @@ def main() -> None:
    validate(module.compile())


+@pytest.mark.skip("Not yet supported")


This worked before, but can no longer compile this since TypeApply is gone. We can recover this test case by having smarter type arg inference in the future

You mean backwards from the point of applying f so that we can compile to f = foo[int] if b else bar[int], sort of thing? Yes, that's certainly possible, sounds pretty cool, might even be what we'd call "bidirectional type inference"....! :)

I wonder if there is some syntax we could add in a future PR to let you manually specify those type args and get this example going if type inference doesn't cope and you really, really, want to. But how often does one do this, I don't think it's a priority!

mark-koch · 2024-05-02T14:08:16Z

tests/integration/test_py.py

@@ -142,6 +142,7 @@ def foo(q1: qubit, q2: qubit, q3: qubit) -> tuple[qubit, qubit, qubit]:
    validate(module.compile())


+@pytest.mark.skip("Requires Tket2 upgrade")


Pytket measurements in tket2 now returns a custom Bit type instead of the prelude bool. This means, we need to load the Tket1 extension for validation, but the extension is still at Hugr v0.3

See #202

Tket2 is on the Hugr 0.4 release candidate now, but it still doesn't work. We need to capture the LBit type or write a conversion pass that turns TKET1 measurements into TKET2 measurements...

acl-cqc

Massive PR but generally looks good, thanks Mark :). A lot of small comments and questions and suggestions for future....

guppylang/tys/builtin.py

acl-cqc · 2024-05-08T09:28:41Z

guppylang/tys/const.py

@@ -36,7 +37,7 @@ class ConstValue(Const):

    # Hugr encoding of the value
    # TODO: We might need a Guppy representation of this...
-    value: val.Value
+    value: ops.Const


not ops.Value? (I admit there are a lot of TODOs around here so it might not really matter!)

Yeah, the Pydantic model in the Hugr repo moved values into the ops module

Yeah. I was thinking this should be a Value, not the Const op-wrapper around one. Is this the thing that gets encoded as a TypeArg? Or used as a node in the Hugr? (If ops.Const is correct you might want to update the comment two lines above)

Ohh sorry I misunderstood. Yes, ops.Value is better here

guppylang/tys/ty.py

scripts/generate_schema.py

guppylang/hugr_builder/hugr.py

acl-cqc · 2024-05-08T14:55:50Z

guppylang/hugr_builder/visualise.py

-            </FONT>
-          </TD>
-        </TR>
+        <TR><TD><FONT POINT-SIZE="{fontsize}" FACE="{fontface}"


I'd drop this reindenting-only change, unless there's some reason to keep it

For some reason, the spacing in the visualisation started to look off. Chaning the indentation fixed it

guppylang/prelude/builtins.py

guppylang/tys/ty.py

tests/error/iter_errors/end_missing.py

codecov-commenter · 2024-05-09T11:22:36Z

Codecov Report

Attention: Patch coverage is 96.78899% with 7 lines in your changes are missing coverage. Please review.

Project coverage is 90.56%. Comparing base (f7adb85) to head (7ff1734).

Files	Patch %	Lines
guppylang/compiler/expr_compiler.py	88.00%	3 Missing ⚠️
guppylang/definition/custom.py	83.33%	1 Missing ⚠️
guppylang/hugr_builder/hugr.py	98.38%	1 Missing ⚠️
guppylang/hugr_builder/visualise.py	0.00%	1 Missing ⚠️
guppylang/tys/ty.py	95.83%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #201      +/-   ##
==========================================
- Coverage   90.98%   90.56%   -0.43%     
==========================================
  Files          47       43       -4     
  Lines        4971     4622     -349     
==========================================
- Hits         4523     4186     -337     
+ Misses        448      436      -12

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

acl-cqc · 2024-05-15T16:10:27Z

guppylang/hugr_builder/hugr.py

                input_nodes.append(n)
-            elif isinstance(n.op, ops.Output):
+            elif isinstance(n.op.root, ops.Output):


Good fix there ;-). I wonder about match n.op.root ?

acl-cqc · 2024-05-15T16:16:11Z

guppylang/prelude/_internal.py

+    Hugr always uses a u64 for the value. The interpretation is:
+      - as an unsigned integer, (value mod `2^N`);
+      - as a signed integer, (value mod `2^(N-1) - 2^(N-1)*a`)
+    where `N = 2^log_width` and `a` is the (N-1)th bit of `x` (counting from 0 = least


Quite detailed! I was going to suggest that it'd be easier just to say 'a' is the most significant bit but I realize that there are bits beyond the N. It might still be easier just to start with "take the N least significant bits of the value. If the value is unsigned...."

Or even, "the interpretation is the twos-complement of the N least significant bits".

I mean, this is correct, so ok ;), but also quite hard to follow.

This is copied straight from the Hugr repo:

https://github.com/CQCL/hugr/blob/1c422abd7d3eb0538635fd8a307de80944f93bcf/hugr/src/std_extensions/arithmetic/int_types.rs#L78-L88

acl-cqc

Already APPROVEd but APPROVE again :-)

TBH the thing here that has the worst "code smell" to my mind is DummyOp. But we know that that's a workaround for a not-yet-existing Hugr feature, so we can expect something a bit, ugh. But certainly, any significant further effort on this PR might be better directed towards that ;-)

mark-koch added 3 commits May 2, 2024 14:38

feat: Upgrade Hugr and start using the shared pydantic model

5c47f40

Merge remote-tracking branch 'origin/main' into chore/hugr-upgrade

07d38b2

Fix ruff and mypy issues

db8f3ad

mark-koch commented May 2, 2024

View reviewed changes

Update Hugr ref

11a1619

mark-koch marked this pull request as ready for review May 3, 2024 08:21

mark-koch mentioned this pull request May 3, 2024

Fix Pytket conversion for circuits with measurement #202

Open

mark-koch requested a review from acl-cqc May 7, 2024 09:49

acl-cqc approved these changes May 8, 2024

View reviewed changes

mark-koch added 14 commits May 9, 2024 09:34

Add helper method to construct tys.FunctionType

9177358

Simplify TupleType.to_hugr()

e58ac85

Delete generate_schema.py

99525e3

Fix ConstInt docstring

d0481a4

Add comment to update_op explaining function edges

d270f87

Remove unnecessary order edges

34818c0

Add comment explaining entry BB detection

9812c32

Clarify DummyOp docstring

648027b

Move assert and rename types to hugr_variants

5e1cac2

Fix add_def and add_declare docstrings

17cdd37

Fix root node serialisation

d1720af

Add rows_to_hugr helper and refactor SumType.to_hugr()

11f18c3

Fix add_load_function docstring

5f9dab3

Merge remote-tracking branch 'origin/main' into chore/hugr-upgrade

e8166b2

mark-koch requested a review from acl-cqc May 9, 2024 11:24

Bump to latest Hugr model

99d77c4

mark-koch added 5 commits May 9, 2024 17:30

Assert argument type in _list_to_hugr

fbf1571

Bump to latest Hugr main

39b4380

Merge remote-tracking branch 'origin/main' into chore/hugr-upgrade

e4d570b

Use hugr 0.4 release candidate

95c1095

Bump tket2 and load it for validation

37c5293

acl-cqc reviewed May 15, 2024

View reviewed changes

acl-cqc approved these changes May 15, 2024

View reviewed changes

mark-koch added 5 commits May 16, 2024 09:11

Use ops.Value instead of ops.Const

7083706

Use match statement in to_raw

7b534eb

Merge remote-tracking branch 'origin/main' into chore/hugr-upgrade

0dc5484

Upgrade merged changes

553d11f

Fix mypy

7ff1734

mark-koch merged commit bd7e67a into main May 16, 2024
4 checks passed

mark-koch deleted the chore/hugr-upgrade branch May 16, 2024 08:39

github-actions bot mentioned this pull request May 15, 2024

chore(main): release 0.3.0 #193

Merged

mark-koch mentioned this pull request May 17, 2024

fix: Mypy tket2 error #220

Merged

mark-koch mentioned this pull request May 29, 2024

Replace RawHugr with quantinuum_hugr's version. #173

Closed

ss2165 mentioned this pull request Jun 5, 2024

Use the serialization schema from quantinuum_hugr #177

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: Upgrade Hugr and start using the shared Pydantic model #201

feat: Upgrade Hugr and start using the shared Pydantic model #201

mark-koch commented May 2, 2024 •

edited

Loading

mark-koch May 2, 2024

mark-koch May 2, 2024

mark-koch May 2, 2024

mark-koch May 2, 2024

acl-cqc May 8, 2024

mark-koch May 9, 2024

acl-cqc May 15, 2024

acl-cqc May 15, 2024

mark-koch May 2, 2024 •

edited

Loading

mark-koch May 15, 2024

acl-cqc left a comment

acl-cqc May 8, 2024

mark-koch May 9, 2024

acl-cqc May 15, 2024

mark-koch May 16, 2024

acl-cqc May 8, 2024

mark-koch May 9, 2024

codecov-commenter commented May 9, 2024 •

edited

Loading

acl-cqc May 15, 2024

acl-cqc May 15, 2024

mark-koch May 16, 2024 •

edited

Loading

acl-cqc left a comment

		@@ -259,6 +261,7 @@ def main() -> None:
		validate(module.compile())


		@pytest.mark.skip("Not yet supported")

		@@ -142,6 +142,7 @@ def foo(q1: qubit, q2: qubit, q3: qubit) -> tuple[qubit, qubit, qubit]:
		validate(module.compile())


		@pytest.mark.skip("Requires Tket2 upgrade")

feat: Upgrade Hugr and start using the shared Pydantic model #201

feat: Upgrade Hugr and start using the shared Pydantic model #201

Conversation

mark-koch commented May 2, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mark-koch May 2, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

acl-cqc left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

codecov-commenter commented May 9, 2024 • edited Loading

Codecov Report

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mark-koch May 16, 2024 • edited Loading

Choose a reason for hiding this comment

acl-cqc left a comment

Choose a reason for hiding this comment

mark-koch commented May 2, 2024 •

edited

Loading

mark-koch May 2, 2024 •

edited

Loading

codecov-commenter commented May 9, 2024 •

edited

Loading

mark-koch May 16, 2024 •

edited

Loading