Fix TikZ copy/paste to preserve parameter metadata across windows by Bumsparkle · Pull Request #433 · zxcalc/zxlive

Bumsparkle · 2026-02-16T13:54:40Z

Issue summary:

Copying/pasting ZX diagrams would drop parameter metadata.

Root cause:

Variable type metadata was not consistently propagated through subgraph copy/merge paths.
Cross-instance transfer goes through TikZ, and symbolic-phase TikZ parsing is fragile in some cases.

Fix summary:

Added explicit variable-type metadata propagation helpers.
Preserved metadata during selection Tikz copy and paste merge paths (including same-instance tab workflows).
Kept TikZ clipboard compatibility while preventing metadata loss in normal ZXLive-to-ZXLive workflows.

Manual verification:

In one window, set a spider phase to a symbolic variable and set variable type in Variables panel.
Copy in window A and paste in window B using TikZ copy/paste (Cmd/Ctrl+Shift+C, Cmd/Ctrl+Shift+V).
Result: diagram pasted across windows and variable metadata/type preserved.
Copy/paste within same instance (tab-to-tab) also preserved metadata/type.

Issue link

Fixes #363

RazinShaikh · 2026-02-16T21:48:45Z

Hi, thanks for the PR. It seems like a reasonable approach to append the additional metadata to tikz for copy/paste across windows. However, when I tested this, it didn't work.

I copied from the left window to the right window in the screenshot below and the variables b and d changed from boolean to parametric.

And for copy-paste between tabs of the same window, the variable viewer is empty so it didn't copy the parameters:

jvdwetering · 2026-02-16T21:58:12Z

Thanks for the implementation!
I'm not so familiar with mime types, would this still make it possible to copy-paste from ZXLive to tikzit and back, or would it no longer be recognized as text?
Also, this would change the behaviour of copy-pasting if I'm not mistaken, making the ctrl-shift-v/c commands no longer necessary (they were the commands that produced tikz instead of the internal graph type).

Bumsparkle · 2026-02-17T14:38:07Z

Fixed my implementation.

Yes, Tikzit copy/paste still works. The clipboard still contains plain TikZ text (mime.setText(...)), so Tikzit continues to recognize it.

Ctrl+Shift+C / Ctrl+Shift+V are still meaningful and kept as TikZ-only copy/paste paths.
Regular Ctrl+C / Ctrl+V now uses internal ZXLive MIME when available (to preserve metadata like variable types) and falls back to TikZ text otherwise, so behavior is improved without removing the TikZ workflow.

Also, I noticed someone else has opened a related PR sooner, so happy to close this one.

lia-approves · 2026-02-17T16:01:57Z

Hi @Bumsparkle, thanks for fixing the 2nd issue Razin pointed out in his most recent comment above, which is that the variable viewer is no longer empty when pasting between tabs of the same window.
The first issue Razin pointed out is still present: when copy pasting between windows (and now also for between tabs of the same window), all Boolean variable types get changed to Parametric variable types - could that be fixed?
We will also check the other submitted PR for the issue but it's the first working one that will be accepted.

Bumsparkle · 2026-02-17T16:31:06Z

Thanks for testing and for the detailed report. I’ve now fixed the remaining issue: boolean variable types are preserved on copy/paste between windows and between tabs (normal Ctrl+C / Ctrl+V), and no longer get converted to parametric.

Bumsparkle · 2026-02-18T10:45:27Z

I’ve resolved the merge conflicts and re-tested the boolean variable type issue: Ctrl+C/Ctrl+V preserves Boolean vs Parametric types across windows and tabs.

RazinShaikh · 2026-02-23T15:22:40Z

Hi @Bumsparkle can you please clean up the PR? There are many changes that are not relevant to the issue; I believe they come from merging incorrectly from the main branch. Can you also make the checks pass?

lia-approves · 2026-02-23T15:27:42Z

I've tested that in ZXLive, copy paste works both with and without Shift, on both different tabs and different windows, preserving the parameter metadata.

The tikz copy (Ctrl/Command + Shift + C in ZXLive) didn't work pasting (Ctrl/Command + V) into Tikzit as-is, which thankfully has a one-line fix in the above commit. Tikzit doesn't preserve the metadata, but that's a feature that could be added to tikzit's parser that I would say isn't necessary for this unitary design bounty.

Bumsparkle · 2026-02-25T11:36:47Z

I’ve finished cleaning up this PR and rebased it so it now only contains the copy/paste variable-type fix (removed unrelated merge-noise files and resolved conflicts).

Could someone with write access please re-run the failed [test (3.11)] job? The failure is in Setup Audio (apt-get install pulseaudio) with Ubuntu mirror 404s, so I think it looks infrastructure-related rather than a code/test failure.

RazinShaikh · 2026-02-25T11:39:39Z

Thanks for cleaning up. We are looking into why setup audio is failing; this is happening across all PRs.

RazinShaikh · 2026-02-25T11:43:54Z

I am looking at the code now and I am a bit confused by the copy_variable_types method. It looks like you are calling this every time a graph is copied. In theory, the code in PyZX should already handle this and include the correct types for variables in the copied graph. Does that mean the copy in PyZX is not working correctly? If that's the case, we should move your copy_variable_types method to PyZX rather than keeping this in ZXLive.
(don't worry, this shouldn't affect the bounty as long as the issue is closed in the end.)

Bumsparkle · 2026-02-25T11:51:44Z

Good point, I’m cautious about over-claiming this as purely a PyZX bug. From what I found, ZXLive’s copy/paste path (subgraph_from_vertices + merge/update), boolean type info was not reliably preserved, which caused variables to appear as parametric after paste. copy_variable_types() is then a defensive fix in ZXLive to keep behaviour stable for users right now.

I agree the cleaner long-term fix is likely to preserve this metadata in PyZX for those graph operations, then simplify/remove the ZXLive workaround.

RazinShaikh · 2026-02-25T12:03:52Z

I see, thanks. Can you document this somewhere in the code so that we can remove the workaround once it has been fixed in PyZX? And do you mind creating an issue in the PyZX repo with a minimal example?

Bumsparkle · 2026-02-25T12:49:04Z

I’ve added a TODO in ZXLive that its a temporary workaround and opened PyZX issue zxcalc/pyzx#408 with a minimal repro.

RazinShaikh · 2026-02-26T15:53:55Z

Hello, it looks like the PyZX issue is already closed (so fast, within 4 hours after you created the issue😅). Do you mind having a look at your PR again and see if the temporary workaround can now be removed? Sorry, I was thinking of leaving the workaround but since PyZX is already fixed, it might be worth cleaning it up before the PR is merged.

Bumsparkle · 2026-02-27T11:03:22Z

Hey, no worries, I've removed the temporary ZXLive-side variable type workaround

RazinShaikh · 2026-02-27T18:27:26Z

zxlive/common.py

+def get_variable_types(graph: GraphT) -> dict[str, bool]:
+    """Returns variable type metadata, falling back to symbolic phase vars."""
+    variable_types = {
+        str(name): bool(graph.var_registry.get_type(name, default=False))
+        for name in graph.var_registry.vars()
+    }
+
+    for v in graph.vertices():
+        phase = graph.phase(v)
+        if not hasattr(phase, "free_vars"):
+            continue
+        try:
+            for var in phase.free_vars():
+                name = str(var)
+                inferred = bool(getattr(var, "is_bool", False))
+                if name not in variable_types:
+                    variable_types[name] = inferred
+                else:
+                    variable_types[name] = variable_types[name] or inferred
+        except Exception:
+            continue
+
+    return variable_types


Is this method necessary now that this is fixed in PyZX? Would something break if you simplified to this?

Suggested change

def get_variable_types(graph: GraphT) -> dict[str, bool]:

"""Returns variable type metadata, falling back to symbolic phase vars."""

variable_types = {

str(name): bool(graph.var_registry.get_type(name, default=False))

for name in graph.var_registry.vars()

}

for v in graph.vertices():

phase = graph.phase(v)

if not hasattr(phase, "free_vars"):

continue

try:

for var in phase.free_vars():

name = str(var)

inferred = bool(getattr(var, "is_bool", False))

if name not in variable_types:

variable_types[name] = inferred

else:

variable_types[name] = variable_types[name] or inferred

except Exception:

continue

return variable_types

def get_variable_types(graph: GraphT) -> dict[str, bool]:

"""Returns variable type metadata."""

variable_types = {

str(name): bool(graph.var_registry.get_type(name, default=False))

for name in graph.var_registry.vars()

}

return variable_types

Don't think anything breaks. Variable types are now taken only from var_registry, which matches PyZX.

RazinShaikh · 2026-02-27T18:42:59Z

zxlive/common.py

+def _coerce_bool(value: object) -> bool:
+    if isinstance(value, bool):
+        return value
+    if isinstance(value, str):
+        return value.strip().lower() in ("true", "1", "yes", "on")
+    return bool(value)


is this method necessary because the from_tikz method will only interact with the metadata encoded by you in _encode_tikz_metadata method? Couldn't you have just done bool(is_bool) instead of _coerce_bool(is_bool)?

You’re right _coerce_bool was just defensive, but as _from_tikz only ever sees data produced by _encode_tikz_metadata, I’ll simplify this to bool(is_bool).

RazinShaikh · 2026-02-27T19:06:18Z

zxlive/common.py

+def normalize_symbolic_phase_types(graph: GraphT) -> None:
+    """Reparse symbolic phases so Var.is_bool matches graph.var_registry."""
+    for v in graph.vertices():
+        if graph.type(v) not in (VertexType.Z, VertexType.X):
+            continue
+        phase = graph.phase(v)
+        if not hasattr(phase, "free_vars"):
+            continue
+        try:
+            if not phase.free_vars():
+                continue
+            graph.set_phase(v, string_to_phase(str(phase), graph))
+        except Exception:
+            continue


Why do we need to parse again and which part of the code makes sure that Var.is_bool matches graph.var_registry? Would graph.rebind_variables_to_registry not suffice instead of normalize_symbolic_phase_types?

Yes graph.rebind_variables_to_registry would work.

RazinShaikh · 2026-02-27T19:08:04Z

zxlive/mainwindow.py

+def _parse_bool(value: object) -> bool:
+    if isinstance(value, bool):
+        return value
+    if isinstance(value, str):
+        return value.strip().lower() in ("true", "1", "yes", "on")
+    return bool(value)


This is the same code is _coerce_bool and as I mentioned there, I am not sure if this is really necessary?

Yep, same story as _coerce_bool, just defensive, I’ve simplified it to bool(v) and removed _parse_bool.

RazinShaikh · 2026-02-27T19:12:05Z

zxlive/mainwindow.py

+        if include_internal:
+            payload = json.dumps({
+                "graph_json": graph.to_json(),
+                "variable_types": get_variable_types(graph),


is it necessary to store variable types seperately for json? I thought json from pyzx already includes the correct variable types

Yes. PyZX’s graph_to_dict already puts variable_types in the graph JSON, and from_json restores the registry. Storing it again in the clipboard was redundant.

RazinShaikh

Thanks for removing the workaround code. Could you please have a look at the above comments?

…stry handling

RazinShaikh · 2026-02-28T12:15:01Z

zxlive/common.py

+def normalize_symbolic_phase_types(graph: GraphT) -> None:
+    """Ensure phase variables are consistent with the graph's var_registry."""
+    # Prefer the library helper if it's available.
+    rebind = getattr(graph, "rebind_variables_to_registry", None)
+    if callable(rebind):
+        rebind()


This method just calls rebind_variables_to_registry in a complicated way. Can you delete this method and replace the instances of normalize_symbolic_phase_types(g) with g.rebind_variables_to_registry?

RazinShaikh · 2026-02-28T12:28:59Z

@Bumsparkle Thanks for the changes. Looking at the code now, I realized that the changes made in this PR are of two types:

fixing to_tikz and from_tikz to preserve metadata
making the zxlive UI work with copy/paste where it calls the correct methods and add variables/types in the variable viewer.

The first one should really be in PyZX rather than ZXLive and ZXLive should expect the PyZX methods to preserve metadata appropriately. It turns out that similar issues were open in PyZX already but I didn't realize that they are related.

I think it would be better to move your code related to the first point to PyZX and that would also end up fixing the existing PyZX issue zxcalc/pyzx#367

Bumsparkle · 2026-02-28T12:34:06Z

@RazinShaikh That makes a lot of sense. I’m going to be away for today and tomorrow morning, but I’ll get started on splitting these changes out this Sunday late afternoon if that's okay?

RazinShaikh · 2026-02-28T12:37:04Z

No worries, I believe the issues need to be closed by Monday end of day for the hackathon

Bumsparkle added 2 commits February 16, 2026 13:37

Fix copy/paste to preserve parameter metadata across windows

b44fb28

Fix mypy rewrite typing and clipboard/proof panel type issues

a628f84

Fix cross-window paste enablement and preserve symbolic tikz phases

a22914a

Fix variable type preservation on paste across windows/tabs

2c7651d

Merge branch 'master' into fix-copy-paste-metadata

29b2b97

to_tikz metadata now works with tikzit's existing parser

85145cc

This was referenced Feb 23, 2026

Bring parameters to Copy/paste #444

Open

Enhance paste_graph method #436

Open

Bumsparkle added 3 commits February 23, 2026 18:08

Revert unrelated mypy/rewrite and style changes.

7eb4472

Fix mypy and flake8 for CI

83b8391

Trim PR scope to copy/paste boolean-type fix only

03ea19e

Document temporary var type workaround pending PyZX fix

f19e390

Remove obsolete var-type workaround after PyZX fix

9929041

RazinShaikh requested changes Feb 27, 2026

View reviewed changes

Bumsparkle added 2 commits February 28, 2026 11:40

Remove redundant variable_types from clipboard and simplify TikZ/regi…

a829445

…stry handling

Merge branch 'master' into fix-copy-paste-metadata

c15f50d

Bumsparkle requested a review from RazinShaikh February 28, 2026 11:50

RazinShaikh requested changes Feb 28, 2026

View reviewed changes

Conversation

Bumsparkle commented Feb 16, 2026

Issue summary:

Root cause:

Fix summary:

Manual verification:

Issue link

Uh oh!

RazinShaikh commented Feb 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jvdwetering commented Feb 16, 2026

Uh oh!

Bumsparkle commented Feb 17, 2026

Uh oh!

lia-approves commented Feb 17, 2026

Uh oh!

Bumsparkle commented Feb 17, 2026

Uh oh!

Bumsparkle commented Feb 18, 2026

Uh oh!

RazinShaikh commented Feb 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lia-approves commented Feb 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Bumsparkle commented Feb 25, 2026

Uh oh!

RazinShaikh commented Feb 25, 2026

Uh oh!

RazinShaikh commented Feb 25, 2026

Uh oh!

Bumsparkle commented Feb 25, 2026

Uh oh!

RazinShaikh commented Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Bumsparkle commented Feb 25, 2026

Uh oh!

RazinShaikh commented Feb 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Bumsparkle commented Feb 27, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

RazinShaikh left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

RazinShaikh commented Feb 28, 2026

Uh oh!

Bumsparkle commented Feb 28, 2026

Uh oh!

RazinShaikh commented Feb 28, 2026

Uh oh!

Reviewers

Assignees

Labels

RazinShaikh commented Feb 16, 2026 •

edited

Loading

RazinShaikh commented Feb 23, 2026 •

edited

Loading

lia-approves commented Feb 23, 2026 •

edited

Loading

RazinShaikh commented Feb 25, 2026 •

edited

Loading

RazinShaikh commented Feb 26, 2026 •

edited

Loading

RazinShaikh left a comment •

edited

Loading