Skip to content

Feature/double named dominance edges in index update #347

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
thomaskrause merged 11 commits into from
Oct 14, 2026
Merged

Feature/double named dominance edges in index update #347

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
2fca938
for performance reasons, it is always best to duplicate named dominan…
MartinKl Oct 9, 2026
6cf783f
update
MartinKl Oct 9, 2026
de484e0
remove unused import
MartinKl Oct 9, 2026
f859801
updated snapshot
MartinKl Oct 9, 2026
24e67c9
Add a benchmark for applying updates to a graph with dominance edges
thomaskrause Oct 12, 2026
2b643ec
Directly write the edge to the component instead of collecting it first
thomaskrause Oct 12, 2026
3f6ffb3
Only add a new edge to the default dominance component if edge is not…
thomaskrause Oct 12, 2026
6dd84c3
Add explicit test that the edges are added
thomaskrause Oct 12, 2026
c6e50e7
Move to member function
thomaskrause Oct 12, 2026
a47fdd6
Merge branch 'apply-update-with-dominance-benchmark' into feature/dou…
thomaskrause Oct 12, 2026
f467758
Only include non-empty graph storages for arity and base edge operato…
thomaskrause Oct 14, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,12 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0

## [Unreleased]

### Added

- `AnnotationComponentType::apply_update_graph_index` makes sure all edges
from named dominance components are also contained in the default
dominance component `Dominance/annis`.

## [4.2.0] - 2026-08-11

### Added
Expand Down
36 changes: 36 additions & 0 deletions graphannis/benches/graphannis.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,8 +13,10 @@ use fake::locales::*;
use graphannis::CorpusStorage;
use graphannis::corpusstorage::ResultOrder;
use graphannis::corpusstorage::{QueryLanguage, SearchQuery};
use graphannis::model::AnnotationComponentType;
use graphannis::update::{GraphUpdate, UpdateEvent};
use std::collections::HashSet;
use std::io::BufReader;
use std::path::Path;
use std::path::PathBuf;

Expand Down Expand Up @@ -429,6 +431,39 @@ fn apply_update_ondisk(bench: &mut Criterion) {
});
}

/// Benchmark a more complex graph with dominance and pointing relations.
fn apply_update_complex_graph(bench: &mut Criterion) {
if CORPUS_STORAGE.is_none() {
return;
}

let cs = CORPUS_STORAGE.as_ref().unwrap();

bench.bench_function("apply_update_complex_graph", move |b| {
cs.create_empty_corpus("apply_update_test_corpus", false)
.unwrap();

let mut input =
BufReader::new(&include_bytes!("../../core/tests/single_sentence.graphml")[..]);
let mut node_updates = GraphUpdate::new();
let mut edge_updates = GraphUpdate::new();
let _config_str = graphannis_core::graph::serialization::graphml::read_graphml::<
AnnotationComponentType,
_,
_,
>(&mut input, &mut node_updates, &mut edge_updates, &|_| {})
.unwrap();

b.iter(|| {
cs.apply_update("apply_update_test_corpus", &mut node_updates)
.unwrap();
cs.apply_update("apply_update_test_corpus", &mut edge_updates)
.unwrap();
});
cs.delete("apply_update_test_corpus").unwrap();
});
}

criterion_group!(name=default; config= Criterion::default().sample_size(50); targets =
deserialize_gum,
find_first_ten_token_gum,
Expand All @@ -441,5 +476,6 @@ criterion_group!(name=default; config= Criterion::default().sample_size(50); tar
criterion_group!(name=apply_update; config= Criterion::default().sample_size(20); targets =
apply_update_inmemory,
apply_update_ondisk,
apply_update_complex_graph,
);
criterion_main!(default, apply_update);
60 changes: 58 additions & 2 deletions graphannis/src/annis/db/aql/model.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,13 +7,13 @@ use facet::Facet;
use graphannis_core::{
annostorage::ValueSearch,
dfs::CycleSafeDFS,
errors::ComponentTypeError,
errors::{ComponentTypeError, Result},
graph::{ANNIS_NS, NODE_TYPE_KEY, storage::union::UnionEdgeContainer},
types::ComponentType,
util::disk_collections::{DEFAULT_BLOCK_CACHE_CAPACITY, DiskMap, EvictionStrategy},
};
use std::{
collections::{BTreeMap, HashMap},
collections::{BTreeMap, HashMap, HashSet},
fmt,
};
use transient_btree_index::BtreeConfig;
Expand Down Expand Up @@ -393,6 +393,60 @@ impl AQLUpdateGraphIndex {
Ok(None)
}
}

/// Make sure all edges in named dominance components are in an unnamed dominance component.
fn add_empty_dominance_edges(
&mut self,
graph: &mut AnnotationGraph,
) -> std::result::Result<(), ComponentTypeError> {
let default_dominance_component = AnnotationComponent::new(
AnnotationComponentType::Dominance,
ANNIS_NS.into(),
"".into(),
);

let existing_dominance_components: Vec<_> =
graph.get_all_components(Some(AnnotationComponentType::Dominance), None);

let (unnamed_dominance_components, named_dominance_componts): (Vec<_>, Vec<_>) =
existing_dominance_components
.into_iter()
.filter(|c| c != &default_dominance_component)
.partition(|c| c.name.is_empty());

let named_dominance_gs: Vec<_> = named_dominance_componts
.into_iter()
.filter_map(|c| graph.get_graphstorage(&c))
.collect();

let unnamed_dominance_gs: Vec<_> = unnamed_dominance_components
.into_iter()
.filter_map(|c| graph.get_graphstorage(&c))
.collect();

let default_dominance_gs = graph.get_or_create_writable(&default_dominance_component)?;
for other_dominance_gs in named_dominance_gs {
for source_node in other_dominance_gs.source_nodes() {
let source = source_node?;
// The ">" operator will search all components that have an empty name.
// We don't need to add an edge if its already part of such a component.
let existing_unnamed_outgoing: Result<Vec<NodeID>> = unnamed_dominance_gs
.iter()
.flat_map(|gs| gs.get_outgoing_edges(source))
.collect();
let existing_unnamed_outgoing =
HashSet::<u64>::from_iter(existing_unnamed_outgoing?);
for target in other_dominance_gs.get_outgoing_edges(source) {
let target = target?;
if !existing_unnamed_outgoing.contains(&target) {
default_dominance_gs.add_edge(Edge { source, target })?;
}
}
}
}

Ok(())
}
}

impl ComponentType for AnnotationComponentType {
Expand Down Expand Up @@ -578,6 +632,8 @@ impl ComponentType for AnnotationComponentType {
}
index.invalid_nodes.compact()?;

index.add_empty_dominance_edges(graph)?;

// Re-index the inherited coverage component.
// To make this operation fast, we need to optimize the order component first
let order_component = AnnotationComponent::new(
Expand Down
198 changes: 198 additions & 0 deletions ...pshots/graphannis__annis__db__aql__model__tests__add_dominance_edges_with_empty_name.snap
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,198 @@
---
source: graphannis/src/annis/db/aql/model/tests.rs
expression: "String::from_utf8_lossy(&output)"
---
<?xml version="1.0" encoding="UTF-8"?>
<graphml>
<key id="k0" for="node" attr.name="syntax::cat" attr.type="string"/>
<key id="k1" for="node" attr.name="annis::doc" attr.type="string"/>
<key id="k2" for="node" attr.name="annis::layer" attr.type="string"/>
<key id="k3" for="node" attr.name="annis::node_type" attr.type="string"/>
<key id="k4" for="node" attr.name="default_ns::pos" attr.type="string"/>
<key id="k5" for="node" attr.name="annis::tok" attr.type="string"/>
<key id="k6" for="node" attr.name="annis::tok-whitespace-before" attr.type="string"/>
<graph edgedefault="directed" parse.order="nodesfirst" parse.nodeids="free" parse.edgeids="canonical">
<node id="single_sentence">
<data key="k3">corpus</data>
</node>
<node id="single_sentence/zossen">
<data key="k1">zossen</data>
<data key="k3">corpus</data>
</node>
<node id="single_sentence/zossen#text">
<data key="k3">datasource</data>
</node>
<node id="single_sentence/zossen#n1">
<data key="k0">ROOT</data>
<data key="k2">syntax</data>
<data key="k3">node</data>
</node>
<node id="single_sentence/zossen#n2">
<data key="k0">S</data>
<data key="k2">syntax</data>
<data key="k3">node</data>
</node>
<node id="single_sentence/zossen#n3">
<data key="k0">NP</data>
<data key="k2">syntax</data>
<data key="k3">node</data>
</node>
<node id="single_sentence/zossen#t1">
<data key="k2">default_layer</data>
<data key="k3">node</data>
<data key="k4">ART</data>
<data key="k5">Die</data>
</node>
<node id="single_sentence/zossen#t2">
<data key="k2">default_layer</data>
<data key="k3">node</data>
<data key="k4">NN</data>
<data key="k5">Jugendlichen</data>
<data key="k6"> </data>
</node>
<node id="single_sentence/zossen#n4">
<data key="k0">PP</data>
<data key="k2">syntax</data>
<data key="k3">node</data>
</node>
<node id="single_sentence/zossen#t3">
<data key="k2">default_layer</data>
<data key="k3">node</data>
<data key="k4">APPR</data>
<data key="k5">in</data>
<data key="k6"> </data>
</node>
<node id="single_sentence/zossen#t4">
<data key="k2">default_layer</data>
<data key="k3">node</data>
<data key="k4">NE</data>
<data key="k5">Zossen</data>
<data key="k6"> </data>
</node>
<node id="single_sentence/zossen#t5">
<data key="k2">default_layer</data>
<data key="k3">node</data>
<data key="k4">VMFIN</data>
<data key="k5">wollen</data>
<data key="k6"> </data>
</node>
<node id="single_sentence/zossen#n5">
<data key="k0">NP</data>
<data key="k2">syntax</data>
<data key="k3">node</data>
</node>
<node id="single_sentence/zossen#t6">
<data key="k2">default_layer</data>
<data key="k3">node</data>
<data key="k4">ART</data>
<data key="k5">ein</data>
<data key="k6"> </data>
</node>
<node id="single_sentence/zossen#t7">
<data key="k2">default_layer</data>
<data key="k3">node</data>
<data key="k4">NN</data>
<data key="k5">Musikcafé</data>
<data key="k6"> </data>
</node>
<node id="single_sentence/zossen#t8">
<data key="k2">default_layer</data>
<data key="k3">node</data>
<data key="k4">$.</data>
<data key="k5">.</data>
<data key="k6"> </data>
</node>
<edge id="e0" source="single_sentence/zossen#n1" target="single_sentence/zossen#n2" label="Dominance/annis/">
</edge>
<edge id="e1" source="single_sentence/zossen#n1" target="single_sentence/zossen#t8" label="Dominance/annis/">
</edge>
<edge id="e2" source="single_sentence/zossen#n2" target="single_sentence/zossen#n3" label="Dominance/annis/">
</edge>
<edge id="e3" source="single_sentence/zossen#n2" target="single_sentence/zossen#t5" label="Dominance/annis/">
</edge>
<edge id="e4" source="single_sentence/zossen#n2" target="single_sentence/zossen#n5" label="Dominance/annis/">
</edge>
<edge id="e5" source="single_sentence/zossen#n3" target="single_sentence/zossen#t1" label="Dominance/annis/">
</edge>
<edge id="e6" source="single_sentence/zossen#n3" target="single_sentence/zossen#t2" label="Dominance/annis/">
</edge>
<edge id="e7" source="single_sentence/zossen#n3" target="single_sentence/zossen#n4" label="Dominance/annis/">
</edge>
<edge id="e8" source="single_sentence/zossen#n4" target="single_sentence/zossen#t3" label="Dominance/annis/">
</edge>
<edge id="e9" source="single_sentence/zossen#n4" target="single_sentence/zossen#t4" label="Dominance/annis/">
</edge>
<edge id="e10" source="single_sentence/zossen#n5" target="single_sentence/zossen#t6" label="Dominance/annis/">
</edge>
<edge id="e11" source="single_sentence/zossen#n5" target="single_sentence/zossen#t7" label="Dominance/annis/">
</edge>
<edge id="e12" source="single_sentence/zossen#n1" target="single_sentence/zossen#n2" label="Dominance/syntax/edge">
</edge>
<edge id="e13" source="single_sentence/zossen#n1" target="single_sentence/zossen#t8" label="Dominance/syntax/edge">
</edge>
<edge id="e14" source="single_sentence/zossen#n2" target="single_sentence/zossen#n3" label="Dominance/syntax/edge">
</edge>
<edge id="e15" source="single_sentence/zossen#n2" target="single_sentence/zossen#t5" label="Dominance/syntax/edge">
</edge>
<edge id="e16" source="single_sentence/zossen#n2" target="single_sentence/zossen#n5" label="Dominance/syntax/edge">
</edge>
<edge id="e17" source="single_sentence/zossen#n3" target="single_sentence/zossen#t1" label="Dominance/syntax/edge">
</edge>
<edge id="e18" source="single_sentence/zossen#n3" target="single_sentence/zossen#t2" label="Dominance/syntax/edge">
</edge>
<edge id="e19" source="single_sentence/zossen#n3" target="single_sentence/zossen#n4" label="Dominance/syntax/edge">
</edge>
<edge id="e20" source="single_sentence/zossen#n4" target="single_sentence/zossen#t3" label="Dominance/syntax/edge">
</edge>
<edge id="e21" source="single_sentence/zossen#n4" target="single_sentence/zossen#t4" label="Dominance/syntax/edge">
</edge>
<edge id="e22" source="single_sentence/zossen#n5" target="single_sentence/zossen#t6" label="Dominance/syntax/edge">
</edge>
<edge id="e23" source="single_sentence/zossen#n5" target="single_sentence/zossen#t7" label="Dominance/syntax/edge">
</edge>
<edge id="e24" source="single_sentence/zossen#t1" target="single_sentence/zossen#t2" label="Ordering/annis/">
</edge>
<edge id="e25" source="single_sentence/zossen#t2" target="single_sentence/zossen#t3" label="Ordering/annis/">
</edge>
<edge id="e26" source="single_sentence/zossen#t3" target="single_sentence/zossen#t4" label="Ordering/annis/">
</edge>
<edge id="e27" source="single_sentence/zossen#t4" target="single_sentence/zossen#t5" label="Ordering/annis/">
</edge>
<edge id="e28" source="single_sentence/zossen#t5" target="single_sentence/zossen#t6" label="Ordering/annis/">
</edge>
<edge id="e29" source="single_sentence/zossen#t6" target="single_sentence/zossen#t7" label="Ordering/annis/">
</edge>
<edge id="e30" source="single_sentence/zossen#t7" target="single_sentence/zossen#t8" label="Ordering/annis/">
</edge>
<edge id="e31" source="single_sentence/zossen" target="single_sentence" label="PartOf/annis/">
</edge>
<edge id="e32" source="single_sentence/zossen#text" target="single_sentence/zossen" label="PartOf/annis/">
</edge>
<edge id="e33" source="single_sentence/zossen#n1" target="single_sentence/zossen#text" label="PartOf/annis/">
</edge>
<edge id="e34" source="single_sentence/zossen#n2" target="single_sentence/zossen#text" label="PartOf/annis/">
</edge>
<edge id="e35" source="single_sentence/zossen#n3" target="single_sentence/zossen#text" label="PartOf/annis/">
</edge>
<edge id="e36" source="single_sentence/zossen#t1" target="single_sentence/zossen#text" label="PartOf/annis/">
</edge>
<edge id="e37" source="single_sentence/zossen#t2" target="single_sentence/zossen#text" label="PartOf/annis/">
</edge>
<edge id="e38" source="single_sentence/zossen#n4" target="single_sentence/zossen#text" label="PartOf/annis/">
</edge>
<edge id="e39" source="single_sentence/zossen#t3" target="single_sentence/zossen#text" label="PartOf/annis/">
</edge>
<edge id="e40" source="single_sentence/zossen#t4" target="single_sentence/zossen#text" label="PartOf/annis/">
</edge>
<edge id="e41" source="single_sentence/zossen#t5" target="single_sentence/zossen#text" label="PartOf/annis/">
</edge>
<edge id="e42" source="single_sentence/zossen#n5" target="single_sentence/zossen#text" label="PartOf/annis/">
</edge>
<edge id="e43" source="single_sentence/zossen#t6" target="single_sentence/zossen#text" label="PartOf/annis/">
</edge>
<edge id="e44" source="single_sentence/zossen#t7" target="single_sentence/zossen#text" label="PartOf/annis/">
</edge>
<edge id="e45" source="single_sentence/zossen#t8" target="single_sentence/zossen#text" label="PartOf/annis/">
</edge>
</graph>
</graphml>
25 changes: 23 additions & 2 deletions graphannis/src/annis/db/aql/model/tests.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,11 +28,11 @@ fn global_stats_token_count() {
assert_eq!(true, graph.global_statistics.is_some());
let global_stats = graph.global_statistics.unwrap();
assert_eq!(true, global_stats.all_token_in_order_component);
assert_matches!(global_stats.corpus_size,
assert_matches!(global_stats.corpus_size,
CorpusSize::Token { base_token_count, segmentation_count }
if base_token_count == 16
&& segmentation_count.len() == 2
&& *segmentation_count.get("diplomatic").unwrap() == 11
&& *segmentation_count.get("diplomatic").unwrap() == 11
&& *segmentation_count.get("norm").unwrap() == 13);
}

Expand Down Expand Up @@ -345,3 +345,24 @@ fn add_token_to_single_sentence() {
.unwrap();
assert_snapshot!(String::from_utf8_lossy(&output));
}

#[test]
fn add_dominance_edges_with_empty_name() {
let content =
&include_bytes!("../../../../../tests/single_sentence_only_named_edges.graphml")[..];
let (graph, _config) =
graphml::import::<AnnotationComponentType, _, _>(content, false, |_| {}).unwrap();

let mut output = Vec::<u8>::default();
graphannis_core::graph::serialization::graphml::export_stable_order(
&graph,
None,
&mut output,
|_| {},
)
.unwrap();

// The output has changed, because new edges have been added.
assert_ne!(content, output);
assert_snapshot!(String::from_utf8_lossy(&output));
}
Loading
Loading