Change temporal values from strings to prop values and add aggregates

[miratepuffin]

API changes

GraphQL: typed values for temporal properties (breaking)

TemporalProperty fields that used to return stringified values now return properly typed PropertyOutput scalars (the same one already used by Property.value), so numbers come back as numbers, booleans as booleans, etc.:

Field	Before	After
values	[String!]!	[PropertyOutput!]!
latest	String	PropertyOutput
at(t)	String	PropertyOutput
unique	[String!]!	[PropertyOutput!]!

No other fields (including the list-of-Property shape of Properties.values / Metadata.values) are changed.

GraphQL: new aggregation fields on TemporalProperty

Parity with the Python TemporalProperty aggregations, exposed as fields on each temporal property (so they compose with window(...) / layer(...) / etc.):

Field	Type	Notes
sum	PropertyOutput	null if dtype isn't additive; strings concatenate in temporal order
mean / average	PropertyOutput	always F64; null if any value can't be converted to f64
count	Int	number of updates
min	PropertyTuple	(time, value) pair; null if dtype isn't comparable
max	PropertyTuple	(time, value) pair; null if dtype isn't comparable
median	PropertyTuple	lower median on even-length inputs

Rust core: aggregations promoted to TemporalPropertyView

sum / min / max / count / mean / average / median previously only existed on the Python wrapper (PyTemporalProp) via private helpers. They've been moved onto the core TemporalPropertyView in raphtory::db::api::properties, so pure-Rust consumers get them directly, and Python + GraphQL now share the one implementation.

Tests

New python/tests/test_base_install/test_graphql/test_gql_temporal_aggregates.py:

• Numeric node property — all aggregations on full timeline, windowed view, and empty-window edge case.
• Float edge property on two layers — all aggregations under full view, per-layer view, and combined layer + window.
• Non-numeric semantics — strings (sum concatenates, mean null, min/max/median lexicographic) and bools (sum/mean null, min/max/median work with False < True).

Fixes #2102

[ljeub-pometry]

do we need the alias?

[miratepuffin]

No, but it has been there for ages and people might be using it

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Rust Benchmark'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 2.

Benchmark suite	Current: 19c6d74	Previous: 9823ef7	Ratio
large/1k random edge additions with numeric string input	1893945 ns/iter (± 272864)	876236 ns/iter (± 86855)	2.16
lotr_graph/num_edges	4 ns/iter (± 0)	0 ns/iter (± 0)	+∞
lotr_graph/num_nodes	140 ns/iter (± 45)	1 ns/iter (± 0)	140
lotr_graph/graph_latest	2 ns/iter (± 0)	0 ns/iter (± 0)	+∞
lotr_graph_materialise/materialize	3241143 ns/iter (± 30322)	1564816 ns/iter (± 35303)	2.07
lotr_graph_window_100/num_edges	25 ns/iter (± 0)	8 ns/iter (± 0)	3.13
lotr_graph_window_100/num_nodes	161 ns/iter (± 57)	5 ns/iter (± 0)	32.20
lotr_graph_window_100/has_node_existing	45 ns/iter (± 1)	22 ns/iter (± 0)	2.05
lotr_graph_window_100_materialise/materialize	3570264 ns/iter (± 8808)	1669150 ns/iter (± 10700)	2.14
lotr_graph_window_10/has_node_existing	143 ns/iter (± 9)	62 ns/iter (± 11)	2.31
lotr_graph_window_10/iterate nodes	30752 ns/iter (± 64)	11339 ns/iter (± 40)	2.71
lotr_graph_window_10_materialise/materialize	1994079 ns/iter (± 9498)	971980 ns/iter (± 4278)	2.05
lotr_graph_subgraph_10pc/has_edge_existing	233 ns/iter (± 11)	93 ns/iter (± 1)	2.51
lotr_graph_subgraph_10pc/num_nodes	148 ns/iter (± 39)	4 ns/iter (± 0)	37
lotr_graph_subgraph_10pc/has_node_existing	99 ns/iter (± 2)	34 ns/iter (± 0)	2.91
lotr_graph_subgraph_10pc/iterate nodes	2549 ns/iter (± 114)	839 ns/iter (± 5)	3.04
lotr_graph_subgraph_10pc_windowed/has_node_existing	137 ns/iter (± 8)	62 ns/iter (± 14)	2.21
lotr_graph_subgraph_10pc_windowed/iterate nodes	4358 ns/iter (± 109)	1365 ns/iter (± 3)	3.19
lotr_graph_window_50_layered/num_edges_temporal	149468 ns/iter (± 1739)	70121 ns/iter (± 7586)	2.13
lotr_graph_window_50_layered/num_nodes	60129 ns/iter (± 670)	21435 ns/iter (± 536)	2.81
lotr_graph_window_50_layered/has_node_existing	891 ns/iter (± 76)	129 ns/iter (± 12)	6.91
lotr_graph_window_50_layered/max_id	68370 ns/iter (± 604)	25556 ns/iter (± 252)	2.68
lotr_graph_window_50_layered/iterate nodes	113580 ns/iter (± 580)	19308 ns/iter (± 47)	5.88
lotr_graph_window_50_layered/iterate edges	171780 ns/iter (± 307)	83616 ns/iter (± 1318)	2.05
lotr_graph_window_50_layered/graph_latest	101737 ns/iter (± 748)	36649 ns/iter (± 916)	2.78
lotr_graph_window_50_layered_materialise/materialize	7073423 ns/iter (± 88907)	3488825 ns/iter (± 24948)	2.03
lotr_graph_persistent_window_50_layered/num_edges_temporal	756950 ns/iter (± 6779)	192686 ns/iter (± 1569)	3.93
lotr_graph_persistent_window_50_layered/num_nodes	81212 ns/iter (± 1435)	31517 ns/iter (± 779)	2.58
lotr_graph_persistent_window_50_layered/has_node_existing	1034 ns/iter (± 149)	174 ns/iter (± 83)	5.94
lotr_graph_persistent_window_50_layered/max_id	90971 ns/iter (± 746)	38024 ns/iter (± 490)	2.39
lotr_graph_persistent_window_50_layered/iterate nodes	155019 ns/iter (± 905)	35886 ns/iter (± 191)	4.32
lotr_graph_persistent_window_50_layered/iterate edges	168339 ns/iter (± 283)	84161 ns/iter (± 596)	2.00
lotr_graph_persistent_window_50_layered/graph_latest	165940 ns/iter (± 2780)	57549 ns/iter (± 4809)	2.88
lotr_graph_persistent_window_50_layered_materialise/materialize	13663360 ns/iter (± 128211)	5298035 ns/iter (± 147912)	2.58
lotr_graph/proto_encode	4791594 ns/iter (± 49562)	1157897 ns/iter (± 73709)	4.14

This comment was automatically generated by workflow using github-action-benchmark.