ctx - expand CeedContextFieldType to support additional field types by ksd-2005 · Pull Request #2008 · CEED/libCEED

ksd-2005 · 2026-08-19T16:49:25Z

Purpose:

Adds seven new CeedContextFieldType values - BYTE, INT8, INT, INT64, SIZE, SCALAR,
and FLOAT - alongside the existing DOUBLE, INT32, and BOOL, with matching
CeedQFunctionContextRegisterCeed* / Set* / Get*Read / Restore*Read functions on both
CeedQFunctionContext and CeedOperator.

As we agreed in the issue discussion, this renumbers the existing CeedContextFieldType enum
(an intentional ABI break, approved ahead of the 1.0 release) and gives every accessor function
a Ceed-prefixed name for the libCEED-typedef'd types, as per @jeremylt 's and @zatkins-dev 's suggestion.

CEED_CONTEXT_FIELD_SCALAR, FLOAT, and DOUBLE are kept as distinct field types.

One thing I want to mention is:
The discussion we had earlier didn't say an exact enum ordering for the final 10-type set - the only ordering anyone said was the case order of
the switch @jeremylt posted (BOOL, BYTE, INT8, INT, INT32, INT64, SIZE, SCALAR, FLOAT, DOUBLE).
I've used

typedef enum {
  /// Byte value, C type of char
  CEED_CONTEXT_FIELD_BYTE = 1,
  /// CeedScalar value
  CEED_CONTEXT_FIELD_SCALAR = 2,
  /// Single precision value
  CEED_CONTEXT_FIELD_FLOAT = 3,
  /// Double precision value
  CEED_CONTEXT_FIELD_DOUBLE = 4,
  ///8 bit integer value
  CEED_CONTEXT_FIELD_INT8 = 5,
  /// CeedInt value
  CEED_CONTEXT_FIELD_INT = 6,
  /// 32 bit integer value
  CEED_CONTEXT_FIELD_INT32 = 7,
  /// 64 bit integer value
  CEED_CONTEXT_FIELD_INT64 = 8,
  /// CeedSize value
  CEED_CONTEXT_FIELD_SIZE = 9,
  /// Boolean value
  CEED_CONTEXT_FIELD_BOOL = 10,
} CeedContextFieldType;

I am happy to change the ordering if anything else was intended.

This PR is currently a draft. The below tasks remain, before it's ready for full review:

Extend tests/t407-qfunction.c to cover all ten field types (currently covers three)
Confirm the enum ordering above
make prove-all clean run including examples/fluids

These will be completed by tomorrow afternoon

Closes: #2004

LLM/GenAI Disclosure:

I used Claude to help me understand the existing CeedQFunctionContext
in interface/ceed-qfunctioncontext.c and interface/ceed-operator.c before
writing this change, and to review my diff against the design decisions from the issue thread -
it pointed out a few bugs (a CeedSize/size_t signature mismatch that failed to compile, a
stale CeedContextFieldTypes[] string table) which I then fixed myself.

ALL CODE SOURCE CHANGES IN THIS PR ARE WRITTEN BY ME.

By submitting this PR, the author certifies to its contents as described by the Developer's Certificate of Origin.

…ieldType Added a Ceed prefix to accessors for the libCEED-typedef'd types.

Updated Clang.jl compat to 0.19 and regenerated bindings Clang.jl 0.16 depends on the deprecated Clang_jll, which no longer resolves in the current registry. Regenerate for CeedContextFieldType expansion.

ksd-2005 · 2026-08-19T17:48:07Z

Please also let me know, if I have to change any coding style( I followed the exact way in which other functions are written in the codebase, but i want to know if I can change anything else), or can improve by making the code concise anywhere, I would like to learn on ways to improve the code.

jeremylt · 2026-08-19T17:52:23Z

Thanks for the PR. It's evening for me here in the EU so I'll look in the morning and let you know!

zatkins-dev · 2026-08-19T21:47:10Z

A note on the ordering: I think I would prefer the order for the switch, it's a bit more intuitive to be. I also think that the enum needs a zero value -- we often use CeedCalloc to get memory and that automatically zeros the allocated memory. Enum values should be valid in that case.

zatkins-dev · 2026-08-19T21:47:46Z

Also, thank you for the very clear explanation of AI use!

zatkins-dev

This is a great start, a couple high-level things on a first look.

zatkins-dev · 2026-08-19T21:50:33Z

+CEED_EXTERN int CeedQFunctionContextSetCeedFloat(CeedQFunctionContext ctx, CeedContextFieldLabel field_label, float *values);
+CEED_EXTERN int  CeedQFunctionContextGetCeedFloatRead(CeedQFunctionContext ctx, CeedContextFieldLabel field_label, size_t *num_values, 
+                                                      const float **values);
+CEED_EXTERN int  CeedQFunctionContextRestoreCeedFloatRead(CeedQFunctionContext ctx, CeedContextFieldLabel field_label, const float **values);


Just to note, only the Ceed types should have the Ceed prefix on the type name for these functions.

For example, CeedScalar should be as you have it, but float should be

Suggested change

CEED_EXTERN int CeedQFunctionContextRestoreCeedFloatRead(CeedQFunctionContext ctx, CeedContextFieldLabel field_label, const float **values);

CEED_EXTERN int CeedQFunctionContextRestoreFloatRead(CeedQFunctionContext ctx, CeedContextFieldLabel field_label, const float **values);

Thats going to be a pretty big diff, so I'll let you handle it locally instead of me suggesting all of them.

Yea, to clarify, I meant only the four types with Ceed in their name should have that added to the signature, so CeedInt, CeedScalar, CeedSize, and CeedInt8

zatkins-dev · 2026-08-19T21:55:01Z

I'm not sure that the Julia changes need to all be here -- some of that should probably be a separate PR.

jeremylt · 2026-08-20T07:28:28Z

 CEED_EXTERN int CeedQFunctionContextRestoreData(CeedQFunctionContext ctx, void *data);
 CEED_EXTERN int CeedQFunctionContextRestoreDataRead(CeedQFunctionContext ctx, void *data);
-CEED_EXTERN int CeedQFunctionContextRegisterDouble(CeedQFunctionContext ctx, const char *field_name, size_t field_offset, size_t num_values,
+CEED_EXTERN int CeedQFunctionContextRegisterCeedByte(CeedQFunctionContext ctx, const char *field_name, size_t field_offset, size_t num_values,


Note, these (and the other fn definitions and declarations) should use the ordering of the enum provided

jeremylt · 2026-08-20T07:28:50Z

 /// @ingroup CeedQFunction
 typedef enum {
+  /// Byte value, C type of char
+  CEED_CONTEXT_FIELD_BYTE = 1,


As Zach said, I'd use the order I provided, starting at 0

jeremylt · 2026-08-20T07:30:11Z

    ccall((:CeedQFunctionSetData, libceed), Cint, (CeedQFunction, Ptr{Cvoid}), qf, data)
 end

+function CeedQFunctionIsImmutable(qf, is_immutable)


there's a bunch of other stuff added here - I'd only add the Julia changes that correspond to C API changes in this PR

jeremylt · 2026-08-20T07:30:59Z

    ccall((:CeedHouseholderApplyQ, libceed), Cint, (Ptr{CeedScalar}, Ptr{CeedScalar}, Ptr{CeedScalar}, CeedTransposeMode, CeedInt, CeedInt, CeedInt, CeedInt, CeedInt), mat_A, mat_Q, tau, t_mode, m, n, k, row, col)
 end

+function CeedMatrixPseudoinverse(ceed, mat, m, n, mat_pinv)


This definitely should not be exposed here

jeremylt · 2026-08-20T07:31:16Z

+function CeedOperatorLinearAssembleQFunctionBuildOrUpdateFallback(op, build_objects, assembled, rstr, request)
+    ccall((:CeedOperatorLinearAssembleQFunctionBuildOrUpdateFallback, libceed), Cint, (CeedOperator, Bool, Ptr{CeedVector}, Ptr{CeedElemRestriction}, Ptr{CeedRequest}), op, build_objects, assembled, rstr, request)
+end
+
+function CeedOperatorAssembleSingle(op, offset, values)


These also don't belong

jeremylt · 2026-08-20T07:31:45Z

+function CeedOperatorHasTensorBases(op, has_tensor_bases)
+    ccall((:CeedOperatorHasTensorBases, libceed), Cint, (CeedOperator, Ptr{Bool}), op, has_tensor_bases)
+end
+
+function CeedOperatorIsImmutable(op, is_immutable)
+    ccall((:CeedOperatorIsImmutable, libceed), Cint, (CeedOperator, Ptr{Bool}), op, is_immutable)


And these are internal utilities that don't belong... really just only add the fn's from this PR and non other to be safe

jeremylt · 2026-08-20T07:35:20Z

+  @ref Backend
+**/
+int CeedQFunctionContextRestoreCeedByteRead(CeedQFunctionContext ctx, CeedContextFieldLabel field_label, const char **values) {
+  CeedCheck(field_label, CeedQFunctionContextReturnCeed(ctx), CEED_ERROR_UNSUPPORTED, "Invalid field label");


These changes emphasize that probably this check should move from being duplicated in all of these functions to being in CeedQFunctionContextRestoreGenericRead

jeremylt · 2026-08-20T07:36:16Z

+  @ref Backend
+**/
+int CeedQFunctionContextSetCeedByte(CeedQFunctionContext ctx, CeedContextFieldLabel field_label, char * values) {
+  CeedCheck(field_label, CeedQFunctionContextReturnCeed(ctx), CEED_ERROR_UNSUPPORTED, "Invalid field label");


same comment here, all these replicated checks probably should go to CeedQFunctionContextSetGeneric

ksd-2005 added 2 commits August 19, 2026 21:46

Added BYTE, INT8, INT, INT64, SIZE, SCALAR, and FLOAT to CeedContextF…

7143c6b

…ieldType Added a Ceed prefix to accessors for the libCEED-typedef'd types.

julia - regenerate bindings for expanded CeedContextFieldType

d9f15c4

Updated Clang.jl compat to 0.19 and regenerated bindings Clang.jl 0.16 depends on the deprecated Clang_jll, which no longer resolves in the current registry. Regenerate for CeedContextFieldType expansion.

zatkins-dev reviewed Aug 19, 2026

View reviewed changes

jeremylt reviewed Aug 20, 2026

View reviewed changes

	CEED_EXTERN int CeedQFunctionContextRestoreCeedFloatRead(CeedQFunctionContext ctx, CeedContextFieldLabel field_label, const float **values);
	CEED_EXTERN int CeedQFunctionContextRestoreFloatRead(CeedQFunctionContext ctx, CeedContextFieldLabel field_label, const float **values);

Uh oh!

Conversation

ksd-2005 commented Aug 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ksd-2005 commented Aug 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jeremylt commented Aug 19, 2026

Uh oh!

zatkins-dev commented Aug 19, 2026

Uh oh!

zatkins-dev commented Aug 19, 2026

Uh oh!

zatkins-dev left a comment

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!

zatkins-dev commented Aug 19, 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!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

ksd-2005 commented Aug 19, 2026 •

edited

Loading

ksd-2005 commented Aug 19, 2026 •

edited

Loading