feat: conventional results post processing #593

ss2165 · 2024-10-24T15:53:08Z

Post processing of result streams, for converting to traditional distributions over bitstrings if some tagging convention is used, including conversion to a pytket BackendResult.

Convention

Tags are assumed to be a name of a bit register unless they fit the regex pattern ^([a-z][\w_]*)\[(\d+)\]$ (like reg[12]) in which case they are assumed to refer to the nth element of a bit register.

For results of the form

result("<register>", value)

value can be {0, 1}, wherein the register is assumed to be length 1, or lists over those values, wherein the list is taken to be the value of the entire register.

For results of the form

result("<register>[n]", value)

value can only be {0, 1}.
The register is assumed to be at least n+1 in size and unset elements are assumed to be 0.

Subsequent writes to the same register/element in the same shot will overwrite.

To convert to a BackendResult all registers must be present in all shots, and register sizes cannot change between shots.

TODO

Clear documentation of the above conventions and examples.

codecov-commenter · 2024-10-24T15:56:44Z

Codecov Report

Attention: Patch coverage is 97.33333% with 2 lines in your changes missing coverage. Please review.

Project coverage is 91.65%. Comparing base (242fa44) to head (a1ee324).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
guppylang/hresult.py	97.33%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #593      +/-   ##
==========================================
+ Coverage   91.58%   91.65%   +0.07%     
==========================================
  Files          59       60       +1     
  Lines        6311     6389      +78     
==========================================
+ Hits         5780     5856      +76     
- Misses        531      533       +2

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

qartik · 2024-10-24T17:38:26Z

tests/test_hresult.py

+        ("as3ABdfAB[2]", ("as3ABdfAB", 2)),
+    ],
+)
+def test_reg_index_pattern_match(identifier, match: tuple[str, int] | None):


It wasn't immediately obvious to me what the purpose of this test was, best to add a docstring.

_s34fd_fd[12] should perhaps be accepted? Do we explicitly forbid register names starting with an underscore?

AFAIK this pattern matches the QASM2 allowed names

tests/test_hresult.py

guppylang/hresult.py

qartik · 2024-10-24T18:12:39Z

guppylang/hresult.py

+For results of the form ``` result("<register>", value) ``` `value` can be `{0, 1, True,
+False}`, wherein the register is assumed to be length 1, or lists over those values,
+wherein the list is taken to be the value of the entire register.
+
+For results of the form ``` result("<register>[n]", value) ``` `value` can only be `{0,
+1, True, False}`. The register is assumed to be at least `n+1` in size and unset


Is there a great benefit in supporting both boolean literals and bits 0 and 1? From the hardware perspective, we probably only want bits.

qartik · 2024-10-25T12:35:29Z

guppylang/hresult.py

+    from pytket.backends.backendresult import BackendResult
+
+#: Primitive data types that can be returned by a result
+DataPrimitive = int | float | bool


I am confused by the spec description above:

result("<register>", value) value can be {0, 1}

and the code here.

Are floats supported or not? I'd imagine, if guppy says result("tag_for_pi", pi), we can return an approx value of pi as part of the results.

Or is this PR limited to bitstrings? If so, why include float?

float is allowed as a result type, it is not supported under the "register bitstring convention" adopted as part of this PR. The HResult and Shots classes faithfully capture all types, it is only the post-processing/conversions that are limited.

qartik

LGTM

mark-koch · 2024-10-29T11:03:16Z

guppylang/hresult.py

+    def to_register_bits(self) -> dict[str, str]:
+        """Convert results to a dictionary of register bit values."""


Let's say I do

results = Results() results.append("qs[0]", 1) results.append("qs", 0) results.append("qs[1]", 1) results.to_register_bits()

The expected output should be {"qs": "01"}, right? I think this works since dicts are iterated in insertion order and HResult.as_dict keeps the order of entries. You could consider using OrderedDict to make this invariant explicit?

yes it relies on insertion order preservation guaranteed from python 3.7 onwards.
OrderedDict is more useful for things that require reordering these days so I will
just add a comment.

mark-koch · 2024-10-29T11:04:30Z

guppylang/hresult.py

+class HShots:
+    """Results accumulated over multiple shots."""
+
+    results: list[HResult] = field(default_factory=list)


No need for the default_factory since you have a custom __init__

mark-koch · 2024-10-29T11:10:20Z

guppylang/hresult.py

+            ) from e
+        counts = self.register_bitstrings(strict_lengths=True, strict_names=True)
+        reg_sizes: dict[str, int] = {
+            reg: len(next(iter(counts[reg]), "")) for reg in counts


Why is the "" default in next needed? Shouldn't all registers be populated with at least one value?

yes, it is just there to avoid having to handle unexpected stopiteration errors induced by changes to other code. It is safe to assume unpopulated registers are size 0

🤖 I have created a release *beep* *boop* --- ## [0.13.0](v0.12.2...v0.13.0) (2024-11-12) ### ⚠ BREAKING CHANGES * `prelude` module renamed to `std` ### Features * add `qubit` discard/measure methods ([#580](#580)) ([242fa44](242fa44)) * Add `SizedIter` wrapper type ([#611](#611)) ([2e9da6b](2e9da6b)) * conventional results post processing ([#593](#593)) ([db96224](db96224)) * Improve compiler diagnostics ([#547](#547)) ([90d465d](90d465d)), closes [#551](#551) [#553](#553) [#586](#586) [#588](#588) [#587](#587) [#590](#590) [#600](#600) [#601](#601) [#606](#606) * restrict result tag sizes to 256 bytes ([#596](#596)) ([4e8e00f](4e8e00f)), closes [#595](#595) ### Bug Fixes * Mock guppy decorator during sphinx builds ([#622](#622)) ([1cccc04](1cccc04)) ### Documentation * Add DEVELOPMENT.md ([#584](#584)) ([1d29d39](1d29d39)) * Fix docs build ([#639](#639)) ([bd6011c](bd6011c)) ### Miscellaneous Chores * Manually set last release commit ([#643](#643)) ([b2d569b](b2d569b)) ### Code Refactoring * rename prelude to std ([#642](#642)) ([1a68e8e](1a68e8e)) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please). --------- Co-authored-by: Agustín Borgna <[email protected]>

first standalone results code

864f61d

ss2165 added 4 commits October 24, 2024 17:44

renamed

4c0bf2e

allow initialisation from tagged streams

329ac82

docstring

6000741

mypy fixes

c4fe5ff

ss2165 changed the title ~~rfc: conventional results post processing~~ feat: conventional results post processing Oct 24, 2024

ss2165 marked this pull request as ready for review October 24, 2024 17:16

ss2165 requested a review from a team as a code owner October 24, 2024 17:16

ss2165 requested a review from qartik October 24, 2024 17:16

qartik reviewed Oct 24, 2024

View reviewed changes

ss2165 added 3 commits October 25, 2024 11:37

remove redundant casts

dcda153

better test docs

6741c76

remove handling of booleans, simplify to bit chars

22469b5

ss2165 requested a review from qartik October 25, 2024 11:04

ss2165 added 2 commits October 25, 2024 12:32

simplify cast

33fa156

change error message

667f711

qartik reviewed Oct 25, 2024

View reviewed changes

qartik approved these changes Oct 25, 2024

View reviewed changes

refactor: rename Shots -> HShots

f306a74

ss2165 requested a review from mark-koch October 29, 2024 10:24

mark-koch approved these changes Oct 29, 2024

View reviewed changes

minor review changes

a1ee324

ss2165 enabled auto-merge October 29, 2024 11:27

ss2165 added this pull request to the merge queue Oct 29, 2024

Merged via the queue into main with commit db96224 Oct 29, 2024
3 checks passed

ss2165 deleted the ss/results branch October 29, 2024 11:34

hugrbot mentioned this pull request Oct 29, 2024

chore(main): release 0.13.0 #518

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: conventional results post processing #593

feat: conventional results post processing #593

ss2165 commented Oct 24, 2024 •

edited

Loading

codecov-commenter commented Oct 24, 2024 •

edited

Loading

qartik Oct 24, 2024 •

edited

Loading

ss2165 Oct 25, 2024

qartik Oct 24, 2024

qartik Oct 25, 2024 •

edited

Loading

ss2165 Oct 25, 2024

qartik left a comment

mark-koch Oct 29, 2024

ss2165 Oct 29, 2024

mark-koch Oct 29, 2024

mark-koch Oct 29, 2024

ss2165 Oct 29, 2024

		def to_register_bits(self) -> dict[str, str]:
		"""Convert results to a dictionary of register bit values."""

feat: conventional results post processing #593

feat: conventional results post processing #593

Conversation

ss2165 commented Oct 24, 2024 • edited Loading

Convention

TODO

codecov-commenter commented Oct 24, 2024 • edited Loading

Codecov Report

qartik Oct 24, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

qartik Oct 25, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

qartik 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

ss2165 commented Oct 24, 2024 •

edited

Loading

codecov-commenter commented Oct 24, 2024 •

edited

Loading

qartik Oct 24, 2024 •

edited

Loading

qartik Oct 25, 2024 •

edited

Loading