FFI: get scalars from arrays

[myrrc]

• Add vx_array_get_scalar method for retrieving vx_scalar.
• Add scalar's primitive, utf8, bool, and binary getters generated from a macro.
• Replace decimal scalar creation functions with a macro.
• Remove vx_array's primitive getters in favor of obtaining a vx_scalar.
• Add C++ API support for getting array's Scalar at index.
  Error propagation in C bindings #7483

[github-actions]

Polar Signals Profiling Results

Latest Run

Status	Commit	Job	Attempt	Link
🟢 Done	b1ac575		1	Explore Profiling Data

Previous Runs (1)

Status	Commit	Job	Attempt	Link
🟢 Done	3945f76		1	Explore Profiling Data

---

Powered by Polar Signals Cloud

[github-actions]

Verdict: No clear signal (low confidence)
Attributed Vortex impact: +2.5%
Engines: DataFusion Likely regression (+11.5%, medium confidence) · DuckDB No clear signal (-5.8%, low confidence)
Vortex (geomean): 0.954x ➖
Parquet (geomean): 0.942x ➖
Shifts: Parquet (control) -5.8% · Median polish -4.3%

How to read Verdict and Engines

• Verdict: Overall PR-level signal after subtracting baseline drift estimated from Parquet control rows. It can be Likely improvement, Likely regression, or No clear signal.
• Engines: Per-engine attribution. DataFusion is compared against DataFusion/Parquet controls; DuckDB is compared against DuckDB/Parquet controls. This answers whether each engine improved or regressed independently.
• Confidence: Based on directional consistency, share of rows above the noise floor, and control-run noise.

---

datafusion / vortex-file-compressed (1.004x ➖, 0↑ 0↓)

name	PR b1ac575 (ns)	base 037b1a8 (ns)	ratio (PR/base)
vortex_q00/datafusion:vortex-file-compressed	9774512	10016685	0.98
vortex_q01/datafusion:vortex-file-compressed	6448746	6241940	1.03

datafusion / parquet (0.901x ➖, 1↑ 0↓)

name	PR b1ac575 (ns)	base 037b1a8 (ns)	ratio (PR/base)
vortex_q00/datafusion:parquet	20056827	21628324	0.93
vortex_q01/datafusion:parquet 🚀	4574698	5231454	0.87

duckdb / vortex-file-compressed (0.929x ➖, 0↑ 0↓)

name	PR b1ac575 (ns)	base 037b1a8 (ns)	ratio (PR/base)
vortex_q00/duckdb:vortex-file-compressed	10144766	10577988	0.96
vortex_q01/duckdb:vortex-file-compressed	5933116	6588883	0.90

duckdb / parquet (0.986x ➖, 0↑ 0↓)

name	PR b1ac575 (ns)	base 037b1a8 (ns)	ratio (PR/base)
vortex_q00/duckdb:parquet	23406979	23594326	0.99
vortex_q01/duckdb:parquet	9363326	9549441	0.98

---

No file size changes detected.

[myrrc]

I've checked again, and I think we want a more gradual change.

We do want to be able to get a scalar for an array, but we can't replace current getters with an intermediate vx_scalar:

1. for bool, we would pay the price of allocating a scalar for every item. This is non-performant, but acceptable, and can be solved by ffi: create bool arrays #8899
2. For strings, however, we can't have an intermediate Scalar because the vx_view's now borrow from a scalar, and C++ api's bulk access would become UB.

I've removed the primitive accessors here.

[joseph-isaacs]

for bool, we would pay the price of allocating a scalar for every item. This is non-performant, but acceptable, and can be solved by #8899
For strings, however, we can't have an intermediate Scalar because the vx_view's now borrow from a scalar, and C++ api's bulk access would become UB.

Can you point to the functions you mean here

[codspeed-hq]

Merging this PR will improve performance by 47.73%

⚡ 1 improved benchmark
✅ 1829 untouched benchmarks
⏩ 43 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	decompress[u64, (10000, 1024)]	175 µs	118.4 µs	+47.73%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing myrrc/ffi-getter-session (901bbe2) with develop (3884b98)}

Footnotes

1. 43 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩