Performance report — pg.Object vs dataclass / pydantic

Active core: rust (native pygx_core when rust). Median ns/op (testing/benchmarks/_bench.py auto-scaling timer); lower is better; — = no equivalent. Numbers are machine-specific (Apple Silicon here) — read ratios, not absolutes.

Regenerate: PYGX_CORE=rust .venv-coredev/bin/python testing/benchmarks/perf_report_bench.py

Notes: 8 fields + 4 kwargs adds 4 undeclared kwargs at construction. pygx Objects (topo=False and topo=True) and dataclass are typed/closed and reject undeclared kwargs by design, so construct / from-json show —; only pydantic (extra="allow") absorbs them. (Read-only ops for that shape run on the 8 declared fields.) symbolic children nests a 2-field object as a field. to/from-json is not identical work across columns: pygx emits a _type tag for round-trippable typed reconstruction (and dispatches on it when loading), whereas dataclass asdict / pydantic model_dump emit bare dicts — extra work attributed to pygx. Class-level utilities (register/load/save/...) and on_sym_* hooks are excluded (not per-instance ops).

Section A — all public operations (3 scalar fields)

Operation	topo=False	topo=True	dataclass	pydantic
Lifecycle (dataclass / pydantic analog)
construct (kwargs)	295.0 ns	283.6 ns	194.5 ns	520.4 ns
attr get	44.6 ns	45.1 ns	39.5 ns	45.2 ns
attr set	136.6 ns	138.5 ns	52.1 ns	142.0 ns
eq (==, same)	314.3 ns	393.7 ns	134.0 ns	295.6 ns
hash (frozen)	457.2 ns	456.5 ns	96.4 ns	109.2 ns
repr	574.5 ns	581.6 ns	324.0 ns	1.08 us
str	581.8 ns	591.7 ns	323.8 ns	937.6 ns
clone (shallow)	507.0 ns	540.9 ns	647.7 ns	599.2 ns
clone (deep)	954.5 ns	963.8 ns	2.31 us	1.73 us
to dict/json	409.5 ns	393.8 ns	631.0 ns	526.9 ns
to json str	2.27 us	2.28 us	1.51 us	609.8 ns
from dict/json	990.3 ns	1.03 us	197.9 ns	669.8 ns
from json str	3.58 us	3.53 us	—	535.4 ns
Symbolic sym_* (no analog)
sym_clone	527.7 ns	566.0 ns	—	—
sym_rebind	911.7 ns	880.0 ns	—	—
sym_seal / unseal	91.6 ns	91.6 ns	—	—
sym_eq	83.1 ns	83.0 ns	—	—
sym_ne	114.9 ns	113.7 ns	—	—
sym_lt	1.22 us	1.24 us	—	—
sym_gt	1.45 us	1.45 us	—	—
sym_hash	390.6 ns	394.9 ns	—	—
sym_missing	575.0 ns	578.3 ns	—	—
sym_nondefault	2.06 us	2.07 us	—	—
sym_keys	232.8 ns	233.3 ns	—	—
sym_values	231.8 ns	232.4 ns	—	—
sym_items	257.8 ns	256.7 ns	—	—
sym_hasattr	169.5 ns	167.0 ns	—	—
sym_getattr	185.2 ns	186.3 ns	—	—
sym_get	444.5 ns	446.3 ns	—	—
sym_has	277.4 ns	283.6 ns	—	—
sym_contains	508.1 ns	516.0 ns	—	—
sym_jsonify	1.03 us	1.30 us	—	—
sym_descendants	1.10 us	1.11 us	—	—
sym_attr_field	172.2 ns	175.2 ns	—	—
format(compact)	407.9 ns	405.3 ns	—	—
query (pg.query)	6.07 us	6.17 us	—	—
Properties sym_* (no analog)
sym_partial	315.5 ns	308.6 ns	—	—
sym_abstract	516.3 ns	515.5 ns	—	—
sym_puresymbolic	184.5 ns	184.7 ns	—	—
sym_sealed	97.8 ns	97.3 ns	—	—
topo_root	—	199.0 ns	—	—
sym_origin	97.6 ns	97.0 ns	—	—
sym_init_args	1.15 us	1.22 us	—	—
topo_path	—	91.1 ns	—	—
topo_parent	—	92.7 ns	—	—
enable_topo	51.1 ns	52.0 ns	—	—
accessor_writable	98.4 ns	98.6 ns	—	—
allow_partial	98.2 ns	97.7 ns	—	—
is_deterministic	1.29 us	1.29 us	—	—

Section B — key operations across scales

construct

Shape	topo=False	topo=True	dataclass	pydantic
3 fields	289.5 ns	285.7 ns	195.7 ns	518.2 ns
4 fields	329.6 ns	319.8 ns	246.8 ns	570.2 ns
8 fields	500.4 ns	499.1 ns	531.5 ns	877.1 ns
16 fields	840.0 ns	847.4 ns	1.26 us	1.44 us
8 fields + 4 kwargs	—	—	—	1.13 us
container children	742.7 ns	804.1 ns	233.7 ns	775.4 ns
symbolic children	1.24 us	2.79 us	299.3 ns	821.1 ns

attr get

Shape	topo=False	topo=True	dataclass	pydantic
3 fields	39.3 ns	39.1 ns	33.9 ns	39.5 ns
4 fields	39.1 ns	39.1 ns	34.3 ns	40.1 ns
8 fields	38.9 ns	39.2 ns	34.1 ns	39.9 ns
16 fields	39.5 ns	39.3 ns	34.0 ns	40.1 ns
8 fields + 4 kwargs	39.5 ns	40.3 ns	34.4 ns	39.9 ns
container children	39.1 ns	39.4 ns	34.1 ns	39.9 ns
symbolic children	39.5 ns	39.4 ns	34.2 ns	40.1 ns

eq

Shape	topo=False	topo=True	dataclass	pydantic
3 fields	297.7 ns	380.6 ns	117.5 ns	280.0 ns
4 fields	311.6 ns	433.9 ns	142.1 ns	289.8 ns
8 fields	326.7 ns	584.2 ns	239.0 ns	309.1 ns
16 fields	382.3 ns	932.0 ns	429.1 ns	364.1 ns
8 fields + 4 kwargs	333.2 ns	593.9 ns	238.0 ns	318.7 ns
container children	522.9 ns	518.4 ns	169.4 ns	325.0 ns
symbolic children	921.2 ns	971.6 ns	221.1 ns	598.4 ns

clone

Shape	topo=False	topo=True	dataclass	pydantic
3 fields	511.1 ns	525.0 ns	610.7 ns	592.2 ns
4 fields	525.4 ns	540.9 ns	665.4 ns	600.7 ns
8 fields	549.5 ns	605.8 ns	675.1 ns	638.6 ns
16 fields	591.4 ns	712.8 ns	765.7 ns	715.2 ns
8 fields + 4 kwargs	567.8 ns	623.6 ns	703.3 ns	632.5 ns
container children	555.4 ns	563.9 ns	662.4 ns	622.9 ns
symbolic children	545.2 ns	3.75 us	637.8 ns	600.3 ns

clone/deep

Shape	topo=False	topo=True	dataclass	pydantic
3 fields	905.4 ns	906.4 ns	2.19 us	1.65 us
4 fields	994.9 ns	985.2 ns	2.49 us	1.83 us
8 fields	1.34 us	1.33 us	3.50 us	2.85 us
16 fields	1.95 us	1.94 us	5.27 us	4.70 us
8 fields + 4 kwargs	1.32 us	1.32 us	3.39 us	3.05 us
container children	4.20 us	4.23 us	4.40 us	3.87 us
symbolic children	2.49 us	3.82 us	4.29 us	3.38 us

to json

Shape	topo=False	topo=True	dataclass	pydantic
3 fields	399.1 ns	384.3 ns	631.7 ns	525.5 ns
4 fields	425.6 ns	425.8 ns	733.3 ns	559.3 ns
8 fields	528.5 ns	622.3 ns	1.18 us	720.2 ns
16 fields	773.8 ns	1.02 us	2.05 us	1.04 us
8 fields + 4 kwargs	530.3 ns	619.3 ns	1.18 us	740.7 ns
container children	672.4 ns	9.69 us	1.66 us	796.8 ns
symbolic children	2.25 us	634.9 ns	1.19 us	686.5 ns

from json

Shape	topo=False	topo=True	dataclass	pydantic
3 fields	978.3 ns	1.01 us	200.9 ns	666.7 ns
4 fields	1.17 us	1.13 us	247.6 ns	701.3 ns
8 fields	1.58 us	1.56 us	529.8 ns	967.8 ns
16 fields	2.49 us	2.46 us	1.28 us	1.41 us
8 fields + 4 kwargs	—	—	—	1.15 us
container children	1.77 us	1.84 us	234.3 ns	897.7 ns
symbolic children	2.04 us	3.70 us	299.7 ns	963.5 ns