kreuzberg 4.0.0.rc1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- checksums.yaml +7 -0
- data/.gitignore +8 -0
- data/.rspec +3 -0
- data/.rubocop.yaml +534 -0
- data/Gemfile +9 -0
- data/Gemfile.lock +157 -0
- data/README.md +421 -0
- data/Rakefile +25 -0
- data/Steepfile +47 -0
- data/examples/async_patterns.rb +340 -0
- data/ext/kreuzberg_rb/extconf.rb +35 -0
- data/ext/kreuzberg_rb/native/Cargo.toml +36 -0
- data/ext/kreuzberg_rb/native/README.md +425 -0
- data/ext/kreuzberg_rb/native/build.rs +17 -0
- data/ext/kreuzberg_rb/native/include/ieeefp.h +11 -0
- data/ext/kreuzberg_rb/native/include/msvc_compat/strings.h +14 -0
- data/ext/kreuzberg_rb/native/include/strings.h +20 -0
- data/ext/kreuzberg_rb/native/include/unistd.h +47 -0
- data/ext/kreuzberg_rb/native/src/lib.rs +2939 -0
- data/extconf.rb +28 -0
- data/kreuzberg.gemspec +105 -0
- data/lib/kreuzberg/api_proxy.rb +142 -0
- data/lib/kreuzberg/cache_api.rb +45 -0
- data/lib/kreuzberg/cli.rb +55 -0
- data/lib/kreuzberg/cli_proxy.rb +127 -0
- data/lib/kreuzberg/config.rb +684 -0
- data/lib/kreuzberg/errors.rb +50 -0
- data/lib/kreuzberg/extraction_api.rb +84 -0
- data/lib/kreuzberg/mcp_proxy.rb +186 -0
- data/lib/kreuzberg/ocr_backend_protocol.rb +113 -0
- data/lib/kreuzberg/post_processor_protocol.rb +86 -0
- data/lib/kreuzberg/result.rb +216 -0
- data/lib/kreuzberg/setup_lib_path.rb +79 -0
- data/lib/kreuzberg/validator_protocol.rb +89 -0
- data/lib/kreuzberg/version.rb +5 -0
- data/lib/kreuzberg.rb +82 -0
- data/pkg/kreuzberg-4.0.0.rc1.gem +0 -0
- data/sig/kreuzberg/internal.rbs +184 -0
- data/sig/kreuzberg.rbs +468 -0
- data/spec/binding/cache_spec.rb +227 -0
- data/spec/binding/cli_proxy_spec.rb +87 -0
- data/spec/binding/cli_spec.rb +54 -0
- data/spec/binding/config_spec.rb +345 -0
- data/spec/binding/config_validation_spec.rb +283 -0
- data/spec/binding/error_handling_spec.rb +213 -0
- data/spec/binding/errors_spec.rb +66 -0
- data/spec/binding/plugins/ocr_backend_spec.rb +307 -0
- data/spec/binding/plugins/postprocessor_spec.rb +269 -0
- data/spec/binding/plugins/validator_spec.rb +274 -0
- data/spec/examples.txt +104 -0
- data/spec/fixtures/config.toml +39 -0
- data/spec/fixtures/config.yaml +42 -0
- data/spec/fixtures/invalid_config.toml +4 -0
- data/spec/smoke/package_spec.rb +178 -0
- data/spec/spec_helper.rb +42 -0
- data/vendor/kreuzberg/Cargo.toml +134 -0
- data/vendor/kreuzberg/README.md +175 -0
- data/vendor/kreuzberg/build.rs +460 -0
- data/vendor/kreuzberg/src/api/error.rs +81 -0
- data/vendor/kreuzberg/src/api/handlers.rs +199 -0
- data/vendor/kreuzberg/src/api/mod.rs +79 -0
- data/vendor/kreuzberg/src/api/server.rs +353 -0
- data/vendor/kreuzberg/src/api/types.rs +170 -0
- data/vendor/kreuzberg/src/bin/profile_extract.rs +455 -0
- data/vendor/kreuzberg/src/cache/mod.rs +1143 -0
- data/vendor/kreuzberg/src/chunking/mod.rs +677 -0
- data/vendor/kreuzberg/src/core/batch_mode.rs +35 -0
- data/vendor/kreuzberg/src/core/config.rs +1032 -0
- data/vendor/kreuzberg/src/core/extractor.rs +903 -0
- data/vendor/kreuzberg/src/core/io.rs +327 -0
- data/vendor/kreuzberg/src/core/mime.rs +615 -0
- data/vendor/kreuzberg/src/core/mod.rs +42 -0
- data/vendor/kreuzberg/src/core/pipeline.rs +906 -0
- data/vendor/kreuzberg/src/embeddings.rs +323 -0
- data/vendor/kreuzberg/src/error.rs +431 -0
- data/vendor/kreuzberg/src/extraction/archive.rs +954 -0
- data/vendor/kreuzberg/src/extraction/docx.rs +40 -0
- data/vendor/kreuzberg/src/extraction/email.rs +854 -0
- data/vendor/kreuzberg/src/extraction/excel.rs +688 -0
- data/vendor/kreuzberg/src/extraction/html.rs +553 -0
- data/vendor/kreuzberg/src/extraction/image.rs +368 -0
- data/vendor/kreuzberg/src/extraction/libreoffice.rs +564 -0
- data/vendor/kreuzberg/src/extraction/mod.rs +77 -0
- data/vendor/kreuzberg/src/extraction/office_metadata/app_properties.rs +398 -0
- data/vendor/kreuzberg/src/extraction/office_metadata/core_properties.rs +247 -0
- data/vendor/kreuzberg/src/extraction/office_metadata/custom_properties.rs +240 -0
- data/vendor/kreuzberg/src/extraction/office_metadata/mod.rs +128 -0
- data/vendor/kreuzberg/src/extraction/pandoc/batch.rs +275 -0
- data/vendor/kreuzberg/src/extraction/pandoc/mime_types.rs +178 -0
- data/vendor/kreuzberg/src/extraction/pandoc/mod.rs +491 -0
- data/vendor/kreuzberg/src/extraction/pandoc/server.rs +496 -0
- data/vendor/kreuzberg/src/extraction/pandoc/subprocess.rs +1188 -0
- data/vendor/kreuzberg/src/extraction/pandoc/version.rs +162 -0
- data/vendor/kreuzberg/src/extraction/pptx.rs +3000 -0
- data/vendor/kreuzberg/src/extraction/structured.rs +490 -0
- data/vendor/kreuzberg/src/extraction/table.rs +328 -0
- data/vendor/kreuzberg/src/extraction/text.rs +269 -0
- data/vendor/kreuzberg/src/extraction/xml.rs +333 -0
- data/vendor/kreuzberg/src/extractors/archive.rs +425 -0
- data/vendor/kreuzberg/src/extractors/docx.rs +479 -0
- data/vendor/kreuzberg/src/extractors/email.rs +129 -0
- data/vendor/kreuzberg/src/extractors/excel.rs +344 -0
- data/vendor/kreuzberg/src/extractors/html.rs +410 -0
- data/vendor/kreuzberg/src/extractors/image.rs +195 -0
- data/vendor/kreuzberg/src/extractors/mod.rs +268 -0
- data/vendor/kreuzberg/src/extractors/pandoc.rs +201 -0
- data/vendor/kreuzberg/src/extractors/pdf.rs +496 -0
- data/vendor/kreuzberg/src/extractors/pptx.rs +234 -0
- data/vendor/kreuzberg/src/extractors/structured.rs +126 -0
- data/vendor/kreuzberg/src/extractors/text.rs +242 -0
- data/vendor/kreuzberg/src/extractors/xml.rs +128 -0
- data/vendor/kreuzberg/src/image/dpi.rs +164 -0
- data/vendor/kreuzberg/src/image/mod.rs +6 -0
- data/vendor/kreuzberg/src/image/preprocessing.rs +417 -0
- data/vendor/kreuzberg/src/image/resize.rs +89 -0
- data/vendor/kreuzberg/src/keywords/config.rs +154 -0
- data/vendor/kreuzberg/src/keywords/mod.rs +237 -0
- data/vendor/kreuzberg/src/keywords/processor.rs +267 -0
- data/vendor/kreuzberg/src/keywords/rake.rs +294 -0
- data/vendor/kreuzberg/src/keywords/types.rs +68 -0
- data/vendor/kreuzberg/src/keywords/yake.rs +163 -0
- data/vendor/kreuzberg/src/language_detection/mod.rs +942 -0
- data/vendor/kreuzberg/src/lib.rs +102 -0
- data/vendor/kreuzberg/src/mcp/mod.rs +32 -0
- data/vendor/kreuzberg/src/mcp/server.rs +1966 -0
- data/vendor/kreuzberg/src/ocr/cache.rs +469 -0
- data/vendor/kreuzberg/src/ocr/error.rs +37 -0
- data/vendor/kreuzberg/src/ocr/hocr.rs +216 -0
- data/vendor/kreuzberg/src/ocr/mod.rs +58 -0
- data/vendor/kreuzberg/src/ocr/processor.rs +847 -0
- data/vendor/kreuzberg/src/ocr/table/mod.rs +4 -0
- data/vendor/kreuzberg/src/ocr/table/tsv_parser.rs +144 -0
- data/vendor/kreuzberg/src/ocr/tesseract_backend.rs +450 -0
- data/vendor/kreuzberg/src/ocr/types.rs +393 -0
- data/vendor/kreuzberg/src/ocr/utils.rs +47 -0
- data/vendor/kreuzberg/src/ocr/validation.rs +206 -0
- data/vendor/kreuzberg/src/pdf/error.rs +122 -0
- data/vendor/kreuzberg/src/pdf/images.rs +139 -0
- data/vendor/kreuzberg/src/pdf/metadata.rs +346 -0
- data/vendor/kreuzberg/src/pdf/mod.rs +50 -0
- data/vendor/kreuzberg/src/pdf/rendering.rs +369 -0
- data/vendor/kreuzberg/src/pdf/table.rs +420 -0
- data/vendor/kreuzberg/src/pdf/text.rs +161 -0
- data/vendor/kreuzberg/src/plugins/extractor.rs +1010 -0
- data/vendor/kreuzberg/src/plugins/mod.rs +209 -0
- data/vendor/kreuzberg/src/plugins/ocr.rs +629 -0
- data/vendor/kreuzberg/src/plugins/processor.rs +641 -0
- data/vendor/kreuzberg/src/plugins/registry.rs +1324 -0
- data/vendor/kreuzberg/src/plugins/traits.rs +258 -0
- data/vendor/kreuzberg/src/plugins/validator.rs +955 -0
- data/vendor/kreuzberg/src/stopwords/mod.rs +1470 -0
- data/vendor/kreuzberg/src/text/mod.rs +19 -0
- data/vendor/kreuzberg/src/text/quality.rs +697 -0
- data/vendor/kreuzberg/src/text/string_utils.rs +217 -0
- data/vendor/kreuzberg/src/text/token_reduction/cjk_utils.rs +164 -0
- data/vendor/kreuzberg/src/text/token_reduction/config.rs +100 -0
- data/vendor/kreuzberg/src/text/token_reduction/core.rs +796 -0
- data/vendor/kreuzberg/src/text/token_reduction/filters.rs +902 -0
- data/vendor/kreuzberg/src/text/token_reduction/mod.rs +160 -0
- data/vendor/kreuzberg/src/text/token_reduction/semantic.rs +619 -0
- data/vendor/kreuzberg/src/text/token_reduction/simd_text.rs +147 -0
- data/vendor/kreuzberg/src/types.rs +873 -0
- data/vendor/kreuzberg/src/utils/mod.rs +17 -0
- data/vendor/kreuzberg/src/utils/quality.rs +959 -0
- data/vendor/kreuzberg/src/utils/string_utils.rs +381 -0
- data/vendor/kreuzberg/stopwords/af_stopwords.json +53 -0
- data/vendor/kreuzberg/stopwords/ar_stopwords.json +482 -0
- data/vendor/kreuzberg/stopwords/bg_stopwords.json +261 -0
- data/vendor/kreuzberg/stopwords/bn_stopwords.json +400 -0
- data/vendor/kreuzberg/stopwords/br_stopwords.json +1205 -0
- data/vendor/kreuzberg/stopwords/ca_stopwords.json +280 -0
- data/vendor/kreuzberg/stopwords/cs_stopwords.json +425 -0
- data/vendor/kreuzberg/stopwords/da_stopwords.json +172 -0
- data/vendor/kreuzberg/stopwords/de_stopwords.json +622 -0
- data/vendor/kreuzberg/stopwords/el_stopwords.json +849 -0
- data/vendor/kreuzberg/stopwords/en_stopwords.json +1300 -0
- data/vendor/kreuzberg/stopwords/eo_stopwords.json +175 -0
- data/vendor/kreuzberg/stopwords/es_stopwords.json +734 -0
- data/vendor/kreuzberg/stopwords/et_stopwords.json +37 -0
- data/vendor/kreuzberg/stopwords/eu_stopwords.json +100 -0
- data/vendor/kreuzberg/stopwords/fa_stopwords.json +801 -0
- data/vendor/kreuzberg/stopwords/fi_stopwords.json +849 -0
- data/vendor/kreuzberg/stopwords/fr_stopwords.json +693 -0
- data/vendor/kreuzberg/stopwords/ga_stopwords.json +111 -0
- data/vendor/kreuzberg/stopwords/gl_stopwords.json +162 -0
- data/vendor/kreuzberg/stopwords/gu_stopwords.json +226 -0
- data/vendor/kreuzberg/stopwords/ha_stopwords.json +41 -0
- data/vendor/kreuzberg/stopwords/he_stopwords.json +196 -0
- data/vendor/kreuzberg/stopwords/hi_stopwords.json +227 -0
- data/vendor/kreuzberg/stopwords/hr_stopwords.json +181 -0
- data/vendor/kreuzberg/stopwords/hu_stopwords.json +791 -0
- data/vendor/kreuzberg/stopwords/hy_stopwords.json +47 -0
- data/vendor/kreuzberg/stopwords/id_stopwords.json +760 -0
- data/vendor/kreuzberg/stopwords/it_stopwords.json +634 -0
- data/vendor/kreuzberg/stopwords/ja_stopwords.json +136 -0
- data/vendor/kreuzberg/stopwords/kn_stopwords.json +84 -0
- data/vendor/kreuzberg/stopwords/ko_stopwords.json +681 -0
- data/vendor/kreuzberg/stopwords/ku_stopwords.json +64 -0
- data/vendor/kreuzberg/stopwords/la_stopwords.json +51 -0
- data/vendor/kreuzberg/stopwords/lt_stopwords.json +476 -0
- data/vendor/kreuzberg/stopwords/lv_stopwords.json +163 -0
- data/vendor/kreuzberg/stopwords/ml_stopwords.json +1 -0
- data/vendor/kreuzberg/stopwords/mr_stopwords.json +101 -0
- data/vendor/kreuzberg/stopwords/ms_stopwords.json +477 -0
- data/vendor/kreuzberg/stopwords/ne_stopwords.json +490 -0
- data/vendor/kreuzberg/stopwords/nl_stopwords.json +415 -0
- data/vendor/kreuzberg/stopwords/no_stopwords.json +223 -0
- data/vendor/kreuzberg/stopwords/pl_stopwords.json +331 -0
- data/vendor/kreuzberg/stopwords/pt_stopwords.json +562 -0
- data/vendor/kreuzberg/stopwords/ro_stopwords.json +436 -0
- data/vendor/kreuzberg/stopwords/ru_stopwords.json +561 -0
- data/vendor/kreuzberg/stopwords/si_stopwords.json +193 -0
- data/vendor/kreuzberg/stopwords/sk_stopwords.json +420 -0
- data/vendor/kreuzberg/stopwords/sl_stopwords.json +448 -0
- data/vendor/kreuzberg/stopwords/so_stopwords.json +32 -0
- data/vendor/kreuzberg/stopwords/st_stopwords.json +33 -0
- data/vendor/kreuzberg/stopwords/sv_stopwords.json +420 -0
- data/vendor/kreuzberg/stopwords/sw_stopwords.json +76 -0
- data/vendor/kreuzberg/stopwords/ta_stopwords.json +129 -0
- data/vendor/kreuzberg/stopwords/te_stopwords.json +54 -0
- data/vendor/kreuzberg/stopwords/th_stopwords.json +118 -0
- data/vendor/kreuzberg/stopwords/tl_stopwords.json +149 -0
- data/vendor/kreuzberg/stopwords/tr_stopwords.json +506 -0
- data/vendor/kreuzberg/stopwords/uk_stopwords.json +75 -0
- data/vendor/kreuzberg/stopwords/ur_stopwords.json +519 -0
- data/vendor/kreuzberg/stopwords/vi_stopwords.json +647 -0
- data/vendor/kreuzberg/stopwords/yo_stopwords.json +62 -0
- data/vendor/kreuzberg/stopwords/zh_stopwords.json +796 -0
- data/vendor/kreuzberg/stopwords/zu_stopwords.json +31 -0
- data/vendor/kreuzberg/tests/api_tests.rs +966 -0
- data/vendor/kreuzberg/tests/archive_integration.rs +543 -0
- data/vendor/kreuzberg/tests/batch_orchestration.rs +542 -0
- data/vendor/kreuzberg/tests/batch_processing.rs +304 -0
- data/vendor/kreuzberg/tests/chunking_offset_demo.rs +92 -0
- data/vendor/kreuzberg/tests/concurrency_stress.rs +509 -0
- data/vendor/kreuzberg/tests/config_features.rs +580 -0
- data/vendor/kreuzberg/tests/config_loading_tests.rs +439 -0
- data/vendor/kreuzberg/tests/core_integration.rs +493 -0
- data/vendor/kreuzberg/tests/csv_integration.rs +424 -0
- data/vendor/kreuzberg/tests/docx_metadata_extraction_test.rs +124 -0
- data/vendor/kreuzberg/tests/email_integration.rs +325 -0
- data/vendor/kreuzberg/tests/error_handling.rs +393 -0
- data/vendor/kreuzberg/tests/format_integration.rs +159 -0
- data/vendor/kreuzberg/tests/helpers/mod.rs +142 -0
- data/vendor/kreuzberg/tests/image_integration.rs +253 -0
- data/vendor/kreuzberg/tests/keywords_integration.rs +479 -0
- data/vendor/kreuzberg/tests/keywords_quality.rs +509 -0
- data/vendor/kreuzberg/tests/mime_detection.rs +428 -0
- data/vendor/kreuzberg/tests/ocr_configuration.rs +510 -0
- data/vendor/kreuzberg/tests/ocr_errors.rs +676 -0
- data/vendor/kreuzberg/tests/ocr_quality.rs +627 -0
- data/vendor/kreuzberg/tests/ocr_stress.rs +469 -0
- data/vendor/kreuzberg/tests/pandoc_integration.rs +503 -0
- data/vendor/kreuzberg/tests/pdf_integration.rs +43 -0
- data/vendor/kreuzberg/tests/pipeline_integration.rs +1412 -0
- data/vendor/kreuzberg/tests/plugin_ocr_backend_test.rs +771 -0
- data/vendor/kreuzberg/tests/plugin_postprocessor_test.rs +561 -0
- data/vendor/kreuzberg/tests/plugin_system.rs +921 -0
- data/vendor/kreuzberg/tests/plugin_validator_test.rs +783 -0
- data/vendor/kreuzberg/tests/registry_integration_tests.rs +607 -0
- data/vendor/kreuzberg/tests/security_validation.rs +404 -0
- data/vendor/kreuzberg/tests/stopwords_integration_test.rs +888 -0
- data/vendor/kreuzberg/tests/test_fastembed.rs +609 -0
- data/vendor/kreuzberg/tests/xlsx_metadata_extraction_test.rs +87 -0
- metadata +471 -0
|
@@ -0,0 +1,921 @@
|
|
|
1
|
+
//! Comprehensive plugin system integration tests.
|
|
2
|
+
//!
|
|
3
|
+
//! Tests plugin registration, discovery, error handling, concurrent access,
|
|
4
|
+
//! and cross-registry interactions for all 4 plugin types.
|
|
5
|
+
|
|
6
|
+
use async_trait::async_trait;
|
|
7
|
+
use kreuzberg::core::config::ExtractionConfig;
|
|
8
|
+
use kreuzberg::plugins::registry::{
|
|
9
|
+
DocumentExtractorRegistry, OcrBackendRegistry, PostProcessorRegistry, ValidatorRegistry,
|
|
10
|
+
};
|
|
11
|
+
use kreuzberg::plugins::{DocumentExtractor, Plugin, PostProcessor, ProcessingStage, Validator};
|
|
12
|
+
use kreuzberg::types::{ExtractionResult, Metadata};
|
|
13
|
+
use kreuzberg::{KreuzbergError, Result};
|
|
14
|
+
use std::sync::Arc;
|
|
15
|
+
|
|
16
|
+
struct FailingExtractor {
|
|
17
|
+
name: String,
|
|
18
|
+
should_fail_init: bool,
|
|
19
|
+
should_fail_extract: bool,
|
|
20
|
+
}
|
|
21
|
+
|
|
22
|
+
impl Plugin for FailingExtractor {
|
|
23
|
+
fn name(&self) -> &str {
|
|
24
|
+
&self.name
|
|
25
|
+
}
|
|
26
|
+
fn version(&self) -> String {
|
|
27
|
+
"1.0.0".to_string()
|
|
28
|
+
}
|
|
29
|
+
fn initialize(&self) -> Result<()> {
|
|
30
|
+
if self.should_fail_init {
|
|
31
|
+
Err(KreuzbergError::Plugin {
|
|
32
|
+
message: "Initialization failed".to_string(),
|
|
33
|
+
plugin_name: self.name.clone(),
|
|
34
|
+
})
|
|
35
|
+
} else {
|
|
36
|
+
Ok(())
|
|
37
|
+
}
|
|
38
|
+
}
|
|
39
|
+
fn shutdown(&self) -> Result<()> {
|
|
40
|
+
Ok(())
|
|
41
|
+
}
|
|
42
|
+
}
|
|
43
|
+
|
|
44
|
+
#[async_trait]
|
|
45
|
+
impl DocumentExtractor for FailingExtractor {
|
|
46
|
+
async fn extract_bytes(&self, _: &[u8], _: &str, _: &ExtractionConfig) -> Result<ExtractionResult> {
|
|
47
|
+
if self.should_fail_extract {
|
|
48
|
+
Err(KreuzbergError::Parsing {
|
|
49
|
+
message: "Extraction failed".to_string(),
|
|
50
|
+
source: None,
|
|
51
|
+
})
|
|
52
|
+
} else {
|
|
53
|
+
Ok(ExtractionResult {
|
|
54
|
+
content: "success".to_string(),
|
|
55
|
+
mime_type: "text/plain".to_string(),
|
|
56
|
+
metadata: Metadata::default(),
|
|
57
|
+
tables: vec![],
|
|
58
|
+
detected_languages: None,
|
|
59
|
+
chunks: None,
|
|
60
|
+
images: None,
|
|
61
|
+
})
|
|
62
|
+
}
|
|
63
|
+
}
|
|
64
|
+
|
|
65
|
+
fn supported_mime_types(&self) -> &[&str] {
|
|
66
|
+
&["text/plain"]
|
|
67
|
+
}
|
|
68
|
+
|
|
69
|
+
fn priority(&self) -> i32 {
|
|
70
|
+
50
|
|
71
|
+
}
|
|
72
|
+
}
|
|
73
|
+
|
|
74
|
+
struct MetadataModifyingProcessor {
|
|
75
|
+
name: String,
|
|
76
|
+
stage: ProcessingStage,
|
|
77
|
+
}
|
|
78
|
+
|
|
79
|
+
impl Plugin for MetadataModifyingProcessor {
|
|
80
|
+
fn name(&self) -> &str {
|
|
81
|
+
&self.name
|
|
82
|
+
}
|
|
83
|
+
fn version(&self) -> String {
|
|
84
|
+
"1.0.0".to_string()
|
|
85
|
+
}
|
|
86
|
+
fn initialize(&self) -> Result<()> {
|
|
87
|
+
Ok(())
|
|
88
|
+
}
|
|
89
|
+
fn shutdown(&self) -> Result<()> {
|
|
90
|
+
Ok(())
|
|
91
|
+
}
|
|
92
|
+
}
|
|
93
|
+
|
|
94
|
+
#[async_trait]
|
|
95
|
+
impl PostProcessor for MetadataModifyingProcessor {
|
|
96
|
+
async fn process(&self, result: &mut ExtractionResult, _: &ExtractionConfig) -> Result<()> {
|
|
97
|
+
result.content.push_str(&format!(" [{}]", self.name));
|
|
98
|
+
Ok(())
|
|
99
|
+
}
|
|
100
|
+
|
|
101
|
+
fn processing_stage(&self) -> ProcessingStage {
|
|
102
|
+
self.stage
|
|
103
|
+
}
|
|
104
|
+
}
|
|
105
|
+
|
|
106
|
+
struct FailingProcessor {
|
|
107
|
+
name: String,
|
|
108
|
+
}
|
|
109
|
+
|
|
110
|
+
impl Plugin for FailingProcessor {
|
|
111
|
+
fn name(&self) -> &str {
|
|
112
|
+
&self.name
|
|
113
|
+
}
|
|
114
|
+
fn version(&self) -> String {
|
|
115
|
+
"1.0.0".to_string()
|
|
116
|
+
}
|
|
117
|
+
fn initialize(&self) -> Result<()> {
|
|
118
|
+
Ok(())
|
|
119
|
+
}
|
|
120
|
+
fn shutdown(&self) -> Result<()> {
|
|
121
|
+
Ok(())
|
|
122
|
+
}
|
|
123
|
+
}
|
|
124
|
+
|
|
125
|
+
#[async_trait]
|
|
126
|
+
impl PostProcessor for FailingProcessor {
|
|
127
|
+
async fn process(&self, _: &mut ExtractionResult, _: &ExtractionConfig) -> Result<()> {
|
|
128
|
+
Err(KreuzbergError::Plugin {
|
|
129
|
+
message: "Processing failed".to_string(),
|
|
130
|
+
plugin_name: self.name.clone(),
|
|
131
|
+
})
|
|
132
|
+
}
|
|
133
|
+
|
|
134
|
+
fn processing_stage(&self) -> ProcessingStage {
|
|
135
|
+
ProcessingStage::Early
|
|
136
|
+
}
|
|
137
|
+
}
|
|
138
|
+
|
|
139
|
+
struct StrictValidator {
|
|
140
|
+
name: String,
|
|
141
|
+
min_length: usize,
|
|
142
|
+
}
|
|
143
|
+
|
|
144
|
+
impl Plugin for StrictValidator {
|
|
145
|
+
fn name(&self) -> &str {
|
|
146
|
+
&self.name
|
|
147
|
+
}
|
|
148
|
+
fn version(&self) -> String {
|
|
149
|
+
"1.0.0".to_string()
|
|
150
|
+
}
|
|
151
|
+
fn initialize(&self) -> Result<()> {
|
|
152
|
+
Ok(())
|
|
153
|
+
}
|
|
154
|
+
fn shutdown(&self) -> Result<()> {
|
|
155
|
+
Ok(())
|
|
156
|
+
}
|
|
157
|
+
}
|
|
158
|
+
|
|
159
|
+
#[async_trait]
|
|
160
|
+
impl Validator for StrictValidator {
|
|
161
|
+
async fn validate(&self, result: &ExtractionResult, _: &ExtractionConfig) -> Result<()> {
|
|
162
|
+
if result.content.len() < self.min_length {
|
|
163
|
+
Err(KreuzbergError::validation(format!(
|
|
164
|
+
"Content too short: {} < {}",
|
|
165
|
+
result.content.len(),
|
|
166
|
+
self.min_length
|
|
167
|
+
)))
|
|
168
|
+
} else {
|
|
169
|
+
Ok(())
|
|
170
|
+
}
|
|
171
|
+
}
|
|
172
|
+
|
|
173
|
+
fn priority(&self) -> i32 {
|
|
174
|
+
50
|
|
175
|
+
}
|
|
176
|
+
}
|
|
177
|
+
|
|
178
|
+
#[test]
|
|
179
|
+
fn test_extractor_registration_failure() {
|
|
180
|
+
let mut registry = DocumentExtractorRegistry::new();
|
|
181
|
+
|
|
182
|
+
let failing_extractor = Arc::new(FailingExtractor {
|
|
183
|
+
name: "failing-extractor".to_string(),
|
|
184
|
+
should_fail_init: true,
|
|
185
|
+
should_fail_extract: false,
|
|
186
|
+
});
|
|
187
|
+
|
|
188
|
+
let result = registry.register(failing_extractor);
|
|
189
|
+
assert!(matches!(result, Err(KreuzbergError::Plugin { .. })));
|
|
190
|
+
}
|
|
191
|
+
|
|
192
|
+
#[tokio::test]
|
|
193
|
+
async fn test_extractor_extraction_failure() {
|
|
194
|
+
let mut registry = DocumentExtractorRegistry::new();
|
|
195
|
+
|
|
196
|
+
let failing_extractor = Arc::new(FailingExtractor {
|
|
197
|
+
name: "failing-extractor".to_string(),
|
|
198
|
+
should_fail_init: false,
|
|
199
|
+
should_fail_extract: true,
|
|
200
|
+
});
|
|
201
|
+
|
|
202
|
+
registry.register(failing_extractor).unwrap();
|
|
203
|
+
|
|
204
|
+
let extractor = registry.get("text/plain").unwrap();
|
|
205
|
+
let config = ExtractionConfig::default();
|
|
206
|
+
let result = extractor.extract_bytes(b"test", "text/plain", &config).await;
|
|
207
|
+
|
|
208
|
+
assert!(matches!(result, Err(KreuzbergError::Parsing { .. })));
|
|
209
|
+
}
|
|
210
|
+
|
|
211
|
+
#[test]
|
|
212
|
+
fn test_extractor_duplicate_registration() {
|
|
213
|
+
let mut registry = DocumentExtractorRegistry::new();
|
|
214
|
+
|
|
215
|
+
let extractor1 = Arc::new(FailingExtractor {
|
|
216
|
+
name: "same-name".to_string(),
|
|
217
|
+
should_fail_init: false,
|
|
218
|
+
should_fail_extract: false,
|
|
219
|
+
});
|
|
220
|
+
|
|
221
|
+
let extractor2 = Arc::new(FailingExtractor {
|
|
222
|
+
name: "same-name".to_string(),
|
|
223
|
+
should_fail_init: false,
|
|
224
|
+
should_fail_extract: false,
|
|
225
|
+
});
|
|
226
|
+
|
|
227
|
+
registry.register(extractor1).unwrap();
|
|
228
|
+
registry.register(extractor2).unwrap();
|
|
229
|
+
|
|
230
|
+
let names = registry.list();
|
|
231
|
+
assert_eq!(names.len(), 1);
|
|
232
|
+
assert!(names.contains(&"same-name".to_string()));
|
|
233
|
+
}
|
|
234
|
+
|
|
235
|
+
#[test]
|
|
236
|
+
fn test_extractor_concurrent_registration() {
|
|
237
|
+
use std::sync::{Arc as StdArc, RwLock};
|
|
238
|
+
use std::thread;
|
|
239
|
+
|
|
240
|
+
let registry = StdArc::new(RwLock::new(DocumentExtractorRegistry::new()));
|
|
241
|
+
let mut handles = vec![];
|
|
242
|
+
|
|
243
|
+
for i in 0..10 {
|
|
244
|
+
let registry_clone = StdArc::clone(®istry);
|
|
245
|
+
let handle = thread::spawn(move || {
|
|
246
|
+
let extractor = Arc::new(FailingExtractor {
|
|
247
|
+
name: format!("extractor-{}", i),
|
|
248
|
+
should_fail_init: false,
|
|
249
|
+
should_fail_extract: false,
|
|
250
|
+
});
|
|
251
|
+
|
|
252
|
+
let mut reg = registry_clone
|
|
253
|
+
.write()
|
|
254
|
+
.expect("Failed to acquire write lock on registry in test");
|
|
255
|
+
reg.register(extractor).unwrap();
|
|
256
|
+
});
|
|
257
|
+
handles.push(handle);
|
|
258
|
+
}
|
|
259
|
+
|
|
260
|
+
for handle in handles {
|
|
261
|
+
handle.join().unwrap();
|
|
262
|
+
}
|
|
263
|
+
|
|
264
|
+
let reg = registry
|
|
265
|
+
.read()
|
|
266
|
+
.expect("Failed to acquire read lock on registry in test");
|
|
267
|
+
assert_eq!(reg.list().len(), 10);
|
|
268
|
+
}
|
|
269
|
+
|
|
270
|
+
#[test]
|
|
271
|
+
fn test_extractor_priority_ordering_complex() {
|
|
272
|
+
let mut registry = DocumentExtractorRegistry::new();
|
|
273
|
+
|
|
274
|
+
struct PriorityExtractor {
|
|
275
|
+
name: String,
|
|
276
|
+
priority: i32,
|
|
277
|
+
}
|
|
278
|
+
|
|
279
|
+
impl Plugin for PriorityExtractor {
|
|
280
|
+
fn name(&self) -> &str {
|
|
281
|
+
&self.name
|
|
282
|
+
}
|
|
283
|
+
fn version(&self) -> String {
|
|
284
|
+
"1.0.0".to_string()
|
|
285
|
+
}
|
|
286
|
+
fn initialize(&self) -> Result<()> {
|
|
287
|
+
Ok(())
|
|
288
|
+
}
|
|
289
|
+
fn shutdown(&self) -> Result<()> {
|
|
290
|
+
Ok(())
|
|
291
|
+
}
|
|
292
|
+
}
|
|
293
|
+
|
|
294
|
+
#[async_trait]
|
|
295
|
+
impl DocumentExtractor for PriorityExtractor {
|
|
296
|
+
async fn extract_bytes(&self, _: &[u8], _: &str, _: &ExtractionConfig) -> Result<ExtractionResult> {
|
|
297
|
+
Ok(ExtractionResult {
|
|
298
|
+
content: "test".to_string(),
|
|
299
|
+
mime_type: "text/plain".to_string(),
|
|
300
|
+
metadata: Metadata::default(),
|
|
301
|
+
tables: vec![],
|
|
302
|
+
detected_languages: None,
|
|
303
|
+
chunks: None,
|
|
304
|
+
images: None,
|
|
305
|
+
})
|
|
306
|
+
}
|
|
307
|
+
fn supported_mime_types(&self) -> &[&str] {
|
|
308
|
+
&["text/plain"]
|
|
309
|
+
}
|
|
310
|
+
fn priority(&self) -> i32 {
|
|
311
|
+
self.priority
|
|
312
|
+
}
|
|
313
|
+
}
|
|
314
|
+
|
|
315
|
+
for priority in [10, 50, 100, 25, 75] {
|
|
316
|
+
let extractor = Arc::new(PriorityExtractor {
|
|
317
|
+
name: format!("priority-{}", priority),
|
|
318
|
+
priority,
|
|
319
|
+
});
|
|
320
|
+
registry.register(extractor).unwrap();
|
|
321
|
+
}
|
|
322
|
+
|
|
323
|
+
let selected = registry.get("text/plain").unwrap();
|
|
324
|
+
assert_eq!(selected.name(), "priority-100");
|
|
325
|
+
assert_eq!(selected.priority(), 100);
|
|
326
|
+
}
|
|
327
|
+
|
|
328
|
+
#[test]
|
|
329
|
+
fn test_extractor_wildcard_vs_exact_priority() {
|
|
330
|
+
let mut registry = DocumentExtractorRegistry::new();
|
|
331
|
+
|
|
332
|
+
let _wildcard = Arc::new(FailingExtractor {
|
|
333
|
+
name: "wildcard-high".to_string(),
|
|
334
|
+
should_fail_init: false,
|
|
335
|
+
should_fail_extract: false,
|
|
336
|
+
});
|
|
337
|
+
|
|
338
|
+
struct WildcardExtractor(FailingExtractor);
|
|
339
|
+
impl Plugin for WildcardExtractor {
|
|
340
|
+
fn name(&self) -> &str {
|
|
341
|
+
self.0.name()
|
|
342
|
+
}
|
|
343
|
+
fn version(&self) -> String {
|
|
344
|
+
self.0.version()
|
|
345
|
+
}
|
|
346
|
+
fn initialize(&self) -> Result<()> {
|
|
347
|
+
Ok(())
|
|
348
|
+
}
|
|
349
|
+
fn shutdown(&self) -> Result<()> {
|
|
350
|
+
Ok(())
|
|
351
|
+
}
|
|
352
|
+
}
|
|
353
|
+
|
|
354
|
+
#[async_trait]
|
|
355
|
+
impl DocumentExtractor for WildcardExtractor {
|
|
356
|
+
async fn extract_bytes(&self, c: &[u8], m: &str, cfg: &ExtractionConfig) -> Result<ExtractionResult> {
|
|
357
|
+
self.0.extract_bytes(c, m, cfg).await
|
|
358
|
+
}
|
|
359
|
+
fn supported_mime_types(&self) -> &[&str] {
|
|
360
|
+
&["text/*"]
|
|
361
|
+
}
|
|
362
|
+
fn priority(&self) -> i32 {
|
|
363
|
+
100
|
|
364
|
+
}
|
|
365
|
+
}
|
|
366
|
+
|
|
367
|
+
let wildcard_arc = Arc::new(WildcardExtractor(FailingExtractor {
|
|
368
|
+
name: "wildcard-high".to_string(),
|
|
369
|
+
should_fail_init: false,
|
|
370
|
+
should_fail_extract: false,
|
|
371
|
+
}));
|
|
372
|
+
|
|
373
|
+
let exact = Arc::new(FailingExtractor {
|
|
374
|
+
name: "exact-low".to_string(),
|
|
375
|
+
should_fail_init: false,
|
|
376
|
+
should_fail_extract: false,
|
|
377
|
+
});
|
|
378
|
+
|
|
379
|
+
registry.register(wildcard_arc).unwrap();
|
|
380
|
+
registry.register(exact).unwrap();
|
|
381
|
+
|
|
382
|
+
let selected = registry.get("text/plain").unwrap();
|
|
383
|
+
assert_eq!(selected.name(), "exact-low");
|
|
384
|
+
}
|
|
385
|
+
|
|
386
|
+
#[test]
|
|
387
|
+
fn test_extractor_empty_mime_type() {
|
|
388
|
+
let registry = DocumentExtractorRegistry::new();
|
|
389
|
+
let result = registry.get("");
|
|
390
|
+
assert!(matches!(result, Err(KreuzbergError::UnsupportedFormat(_))));
|
|
391
|
+
}
|
|
392
|
+
|
|
393
|
+
#[test]
|
|
394
|
+
fn test_extractor_special_characters_mime() {
|
|
395
|
+
let registry = DocumentExtractorRegistry::new();
|
|
396
|
+
let result = registry.get("application/vnd.openxmlformats-officedocument.wordprocessingml.document");
|
|
397
|
+
assert!(matches!(result, Err(KreuzbergError::UnsupportedFormat(_))));
|
|
398
|
+
}
|
|
399
|
+
|
|
400
|
+
#[test]
|
|
401
|
+
fn test_extractor_remove_nonexistent() {
|
|
402
|
+
let mut registry = DocumentExtractorRegistry::new();
|
|
403
|
+
let result = registry.remove("nonexistent");
|
|
404
|
+
assert!(result.is_ok());
|
|
405
|
+
}
|
|
406
|
+
|
|
407
|
+
#[test]
|
|
408
|
+
fn test_extractor_list_after_partial_removal() {
|
|
409
|
+
let mut registry = DocumentExtractorRegistry::new();
|
|
410
|
+
|
|
411
|
+
for i in 0..5 {
|
|
412
|
+
let extractor = Arc::new(FailingExtractor {
|
|
413
|
+
name: format!("extractor-{}", i),
|
|
414
|
+
should_fail_init: false,
|
|
415
|
+
should_fail_extract: false,
|
|
416
|
+
});
|
|
417
|
+
registry.register(extractor).unwrap();
|
|
418
|
+
}
|
|
419
|
+
|
|
420
|
+
registry.remove("extractor-2").unwrap();
|
|
421
|
+
registry.remove("extractor-3").unwrap();
|
|
422
|
+
|
|
423
|
+
let names = registry.list();
|
|
424
|
+
assert_eq!(names.len(), 3);
|
|
425
|
+
assert!(names.contains(&"extractor-0".to_string()));
|
|
426
|
+
assert!(names.contains(&"extractor-1".to_string()));
|
|
427
|
+
assert!(names.contains(&"extractor-4".to_string()));
|
|
428
|
+
}
|
|
429
|
+
|
|
430
|
+
#[tokio::test]
|
|
431
|
+
async fn test_processor_execution_order_within_stage() {
|
|
432
|
+
let mut registry = PostProcessorRegistry::new();
|
|
433
|
+
|
|
434
|
+
let high = Arc::new(MetadataModifyingProcessor {
|
|
435
|
+
name: "high".to_string(),
|
|
436
|
+
stage: ProcessingStage::Early,
|
|
437
|
+
});
|
|
438
|
+
|
|
439
|
+
let medium = Arc::new(MetadataModifyingProcessor {
|
|
440
|
+
name: "medium".to_string(),
|
|
441
|
+
stage: ProcessingStage::Early,
|
|
442
|
+
});
|
|
443
|
+
|
|
444
|
+
let low = Arc::new(MetadataModifyingProcessor {
|
|
445
|
+
name: "low".to_string(),
|
|
446
|
+
stage: ProcessingStage::Early,
|
|
447
|
+
});
|
|
448
|
+
|
|
449
|
+
registry.register(low, 10).unwrap();
|
|
450
|
+
registry.register(high, 100).unwrap();
|
|
451
|
+
registry.register(medium, 50).unwrap();
|
|
452
|
+
|
|
453
|
+
let processors = registry.get_for_stage(ProcessingStage::Early);
|
|
454
|
+
assert_eq!(processors.len(), 3);
|
|
455
|
+
|
|
456
|
+
let mut result = ExtractionResult {
|
|
457
|
+
content: "start".to_string(),
|
|
458
|
+
mime_type: "text/plain".to_string(),
|
|
459
|
+
metadata: Metadata::default(),
|
|
460
|
+
tables: vec![],
|
|
461
|
+
detected_languages: None,
|
|
462
|
+
chunks: None,
|
|
463
|
+
images: None,
|
|
464
|
+
};
|
|
465
|
+
|
|
466
|
+
let config = ExtractionConfig::default();
|
|
467
|
+
for processor in processors {
|
|
468
|
+
processor.process(&mut result, &config).await.unwrap();
|
|
469
|
+
}
|
|
470
|
+
|
|
471
|
+
assert_eq!(result.content, "start [high] [medium] [low]");
|
|
472
|
+
}
|
|
473
|
+
|
|
474
|
+
#[tokio::test]
|
|
475
|
+
async fn test_processor_error_propagation() {
|
|
476
|
+
let mut registry = PostProcessorRegistry::new();
|
|
477
|
+
|
|
478
|
+
let failing = Arc::new(FailingProcessor {
|
|
479
|
+
name: "failing".to_string(),
|
|
480
|
+
});
|
|
481
|
+
|
|
482
|
+
registry.register(failing, 50).unwrap();
|
|
483
|
+
|
|
484
|
+
let processors = registry.get_for_stage(ProcessingStage::Early);
|
|
485
|
+
assert_eq!(processors.len(), 1);
|
|
486
|
+
|
|
487
|
+
let mut result = ExtractionResult {
|
|
488
|
+
content: "test".to_string(),
|
|
489
|
+
mime_type: "text/plain".to_string(),
|
|
490
|
+
metadata: Metadata::default(),
|
|
491
|
+
tables: vec![],
|
|
492
|
+
detected_languages: None,
|
|
493
|
+
chunks: None,
|
|
494
|
+
images: None,
|
|
495
|
+
};
|
|
496
|
+
|
|
497
|
+
let config = ExtractionConfig::default();
|
|
498
|
+
let process_result = processors[0].process(&mut result, &config).await;
|
|
499
|
+
|
|
500
|
+
assert!(matches!(process_result, Err(KreuzbergError::Plugin { .. })));
|
|
501
|
+
}
|
|
502
|
+
|
|
503
|
+
#[test]
|
|
504
|
+
fn test_processor_multiple_stages() {
|
|
505
|
+
let mut registry = PostProcessorRegistry::new();
|
|
506
|
+
|
|
507
|
+
let early = Arc::new(MetadataModifyingProcessor {
|
|
508
|
+
name: "early".to_string(),
|
|
509
|
+
stage: ProcessingStage::Early,
|
|
510
|
+
});
|
|
511
|
+
|
|
512
|
+
let middle = Arc::new(MetadataModifyingProcessor {
|
|
513
|
+
name: "middle".to_string(),
|
|
514
|
+
stage: ProcessingStage::Middle,
|
|
515
|
+
});
|
|
516
|
+
|
|
517
|
+
let late = Arc::new(MetadataModifyingProcessor {
|
|
518
|
+
name: "late".to_string(),
|
|
519
|
+
stage: ProcessingStage::Late,
|
|
520
|
+
});
|
|
521
|
+
|
|
522
|
+
registry.register(early, 50).unwrap();
|
|
523
|
+
registry.register(middle, 50).unwrap();
|
|
524
|
+
registry.register(late, 50).unwrap();
|
|
525
|
+
|
|
526
|
+
assert_eq!(registry.get_for_stage(ProcessingStage::Early).len(), 1);
|
|
527
|
+
assert_eq!(registry.get_for_stage(ProcessingStage::Middle).len(), 1);
|
|
528
|
+
assert_eq!(registry.get_for_stage(ProcessingStage::Late).len(), 1);
|
|
529
|
+
}
|
|
530
|
+
|
|
531
|
+
#[test]
|
|
532
|
+
fn test_processor_registration_failure() {
|
|
533
|
+
struct FailingInitProcessor;
|
|
534
|
+
|
|
535
|
+
impl Plugin for FailingInitProcessor {
|
|
536
|
+
fn name(&self) -> &str {
|
|
537
|
+
"failing-init"
|
|
538
|
+
}
|
|
539
|
+
fn version(&self) -> String {
|
|
540
|
+
"1.0.0".to_string()
|
|
541
|
+
}
|
|
542
|
+
fn initialize(&self) -> Result<()> {
|
|
543
|
+
Err(KreuzbergError::Plugin {
|
|
544
|
+
message: "Init failed".to_string(),
|
|
545
|
+
plugin_name: "failing-init".to_string(),
|
|
546
|
+
})
|
|
547
|
+
}
|
|
548
|
+
fn shutdown(&self) -> Result<()> {
|
|
549
|
+
Ok(())
|
|
550
|
+
}
|
|
551
|
+
}
|
|
552
|
+
|
|
553
|
+
#[async_trait]
|
|
554
|
+
impl PostProcessor for FailingInitProcessor {
|
|
555
|
+
async fn process(&self, _: &mut ExtractionResult, _: &ExtractionConfig) -> Result<()> {
|
|
556
|
+
Ok(())
|
|
557
|
+
}
|
|
558
|
+
fn processing_stage(&self) -> ProcessingStage {
|
|
559
|
+
ProcessingStage::Early
|
|
560
|
+
}
|
|
561
|
+
}
|
|
562
|
+
|
|
563
|
+
let mut registry = PostProcessorRegistry::new();
|
|
564
|
+
let processor = Arc::new(FailingInitProcessor);
|
|
565
|
+
|
|
566
|
+
let result = registry.register(processor, 50);
|
|
567
|
+
assert!(matches!(result, Err(KreuzbergError::Plugin { .. })));
|
|
568
|
+
}
|
|
569
|
+
|
|
570
|
+
#[test]
|
|
571
|
+
fn test_processor_same_priority_same_stage() {
|
|
572
|
+
let mut registry = PostProcessorRegistry::new();
|
|
573
|
+
|
|
574
|
+
let proc1 = Arc::new(MetadataModifyingProcessor {
|
|
575
|
+
name: "processor1".to_string(),
|
|
576
|
+
stage: ProcessingStage::Early,
|
|
577
|
+
});
|
|
578
|
+
|
|
579
|
+
let proc2 = Arc::new(MetadataModifyingProcessor {
|
|
580
|
+
name: "processor2".to_string(),
|
|
581
|
+
stage: ProcessingStage::Early,
|
|
582
|
+
});
|
|
583
|
+
|
|
584
|
+
registry.register(proc1, 50).unwrap();
|
|
585
|
+
registry.register(proc2, 50).unwrap();
|
|
586
|
+
|
|
587
|
+
let processors = registry.get_for_stage(ProcessingStage::Early);
|
|
588
|
+
assert_eq!(processors.len(), 2);
|
|
589
|
+
}
|
|
590
|
+
|
|
591
|
+
#[test]
|
|
592
|
+
fn test_processor_remove_from_specific_stage() {
|
|
593
|
+
let mut registry = PostProcessorRegistry::new();
|
|
594
|
+
|
|
595
|
+
let early = Arc::new(MetadataModifyingProcessor {
|
|
596
|
+
name: "processor".to_string(),
|
|
597
|
+
stage: ProcessingStage::Early,
|
|
598
|
+
});
|
|
599
|
+
|
|
600
|
+
registry.register(early, 50).unwrap();
|
|
601
|
+
assert_eq!(registry.get_for_stage(ProcessingStage::Early).len(), 1);
|
|
602
|
+
|
|
603
|
+
registry.remove("processor").unwrap();
|
|
604
|
+
assert_eq!(registry.get_for_stage(ProcessingStage::Early).len(), 0);
|
|
605
|
+
}
|
|
606
|
+
|
|
607
|
+
#[test]
|
|
608
|
+
fn test_processor_list_across_stages() {
|
|
609
|
+
let mut registry = PostProcessorRegistry::new();
|
|
610
|
+
|
|
611
|
+
for stage in [ProcessingStage::Early, ProcessingStage::Middle, ProcessingStage::Late] {
|
|
612
|
+
let processor = Arc::new(MetadataModifyingProcessor {
|
|
613
|
+
name: format!("{:?}-processor", stage),
|
|
614
|
+
stage,
|
|
615
|
+
});
|
|
616
|
+
registry.register(processor, 50).unwrap();
|
|
617
|
+
}
|
|
618
|
+
|
|
619
|
+
let names = registry.list();
|
|
620
|
+
assert_eq!(names.len(), 3);
|
|
621
|
+
}
|
|
622
|
+
|
|
623
|
+
#[test]
|
|
624
|
+
fn test_processor_shutdown_clears_all_stages() {
|
|
625
|
+
let mut registry = PostProcessorRegistry::new();
|
|
626
|
+
|
|
627
|
+
for stage in [ProcessingStage::Early, ProcessingStage::Middle, ProcessingStage::Late] {
|
|
628
|
+
let processor = Arc::new(MetadataModifyingProcessor {
|
|
629
|
+
name: format!("{:?}-processor", stage),
|
|
630
|
+
stage,
|
|
631
|
+
});
|
|
632
|
+
registry.register(processor, 50).unwrap();
|
|
633
|
+
}
|
|
634
|
+
|
|
635
|
+
registry.shutdown_all().unwrap();
|
|
636
|
+
|
|
637
|
+
assert_eq!(registry.get_for_stage(ProcessingStage::Early).len(), 0);
|
|
638
|
+
assert_eq!(registry.get_for_stage(ProcessingStage::Middle).len(), 0);
|
|
639
|
+
assert_eq!(registry.get_for_stage(ProcessingStage::Late).len(), 0);
|
|
640
|
+
}
|
|
641
|
+
|
|
642
|
+
#[tokio::test]
|
|
643
|
+
async fn test_validator_content_validation() {
|
|
644
|
+
let mut registry = ValidatorRegistry::new();
|
|
645
|
+
|
|
646
|
+
let strict = Arc::new(StrictValidator {
|
|
647
|
+
name: "strict".to_string(),
|
|
648
|
+
min_length: 10,
|
|
649
|
+
});
|
|
650
|
+
|
|
651
|
+
registry.register(strict).unwrap();
|
|
652
|
+
|
|
653
|
+
let validators = registry.get_all();
|
|
654
|
+
assert_eq!(validators.len(), 1);
|
|
655
|
+
|
|
656
|
+
let config = ExtractionConfig::default();
|
|
657
|
+
|
|
658
|
+
let short_result = ExtractionResult {
|
|
659
|
+
content: "short".to_string(),
|
|
660
|
+
mime_type: "text/plain".to_string(),
|
|
661
|
+
metadata: Metadata::default(),
|
|
662
|
+
tables: vec![],
|
|
663
|
+
detected_languages: None,
|
|
664
|
+
chunks: None,
|
|
665
|
+
images: None,
|
|
666
|
+
};
|
|
667
|
+
|
|
668
|
+
let validation = validators[0].validate(&short_result, &config).await;
|
|
669
|
+
assert!(matches!(validation, Err(KreuzbergError::Validation { .. })));
|
|
670
|
+
|
|
671
|
+
let long_result = ExtractionResult {
|
|
672
|
+
content: "this is long enough content".to_string(),
|
|
673
|
+
mime_type: "text/plain".to_string(),
|
|
674
|
+
metadata: Metadata::default(),
|
|
675
|
+
tables: vec![],
|
|
676
|
+
detected_languages: None,
|
|
677
|
+
chunks: None,
|
|
678
|
+
images: None,
|
|
679
|
+
};
|
|
680
|
+
|
|
681
|
+
let validation = validators[0].validate(&long_result, &config).await;
|
|
682
|
+
assert!(validation.is_ok());
|
|
683
|
+
}
|
|
684
|
+
|
|
685
|
+
#[test]
|
|
686
|
+
fn test_validator_priority_ordering() {
|
|
687
|
+
let mut registry = ValidatorRegistry::new();
|
|
688
|
+
|
|
689
|
+
let _high = Arc::new(StrictValidator {
|
|
690
|
+
name: "high-priority".to_string(),
|
|
691
|
+
min_length: 5,
|
|
692
|
+
});
|
|
693
|
+
|
|
694
|
+
struct MediumPriorityValidator;
|
|
695
|
+
impl Plugin for MediumPriorityValidator {
|
|
696
|
+
fn name(&self) -> &str {
|
|
697
|
+
"medium-priority"
|
|
698
|
+
}
|
|
699
|
+
fn version(&self) -> String {
|
|
700
|
+
"1.0.0".to_string()
|
|
701
|
+
}
|
|
702
|
+
fn initialize(&self) -> Result<()> {
|
|
703
|
+
Ok(())
|
|
704
|
+
}
|
|
705
|
+
fn shutdown(&self) -> Result<()> {
|
|
706
|
+
Ok(())
|
|
707
|
+
}
|
|
708
|
+
}
|
|
709
|
+
|
|
710
|
+
#[async_trait]
|
|
711
|
+
impl Validator for MediumPriorityValidator {
|
|
712
|
+
async fn validate(&self, _: &ExtractionResult, _: &ExtractionConfig) -> Result<()> {
|
|
713
|
+
Ok(())
|
|
714
|
+
}
|
|
715
|
+
fn priority(&self) -> i32 {
|
|
716
|
+
50
|
|
717
|
+
}
|
|
718
|
+
}
|
|
719
|
+
|
|
720
|
+
struct LowPriorityValidator;
|
|
721
|
+
impl Plugin for LowPriorityValidator {
|
|
722
|
+
fn name(&self) -> &str {
|
|
723
|
+
"low-priority"
|
|
724
|
+
}
|
|
725
|
+
fn version(&self) -> String {
|
|
726
|
+
"1.0.0".to_string()
|
|
727
|
+
}
|
|
728
|
+
fn initialize(&self) -> Result<()> {
|
|
729
|
+
Ok(())
|
|
730
|
+
}
|
|
731
|
+
fn shutdown(&self) -> Result<()> {
|
|
732
|
+
Ok(())
|
|
733
|
+
}
|
|
734
|
+
}
|
|
735
|
+
|
|
736
|
+
#[async_trait]
|
|
737
|
+
impl Validator for LowPriorityValidator {
|
|
738
|
+
async fn validate(&self, _: &ExtractionResult, _: &ExtractionConfig) -> Result<()> {
|
|
739
|
+
Ok(())
|
|
740
|
+
}
|
|
741
|
+
fn priority(&self) -> i32 {
|
|
742
|
+
10
|
|
743
|
+
}
|
|
744
|
+
}
|
|
745
|
+
|
|
746
|
+
struct HighPriorityValidator;
|
|
747
|
+
impl Plugin for HighPriorityValidator {
|
|
748
|
+
fn name(&self) -> &str {
|
|
749
|
+
"high-priority"
|
|
750
|
+
}
|
|
751
|
+
fn version(&self) -> String {
|
|
752
|
+
"1.0.0".to_string()
|
|
753
|
+
}
|
|
754
|
+
fn initialize(&self) -> Result<()> {
|
|
755
|
+
Ok(())
|
|
756
|
+
}
|
|
757
|
+
fn shutdown(&self) -> Result<()> {
|
|
758
|
+
Ok(())
|
|
759
|
+
}
|
|
760
|
+
}
|
|
761
|
+
|
|
762
|
+
#[async_trait]
|
|
763
|
+
impl Validator for HighPriorityValidator {
|
|
764
|
+
async fn validate(&self, _: &ExtractionResult, _: &ExtractionConfig) -> Result<()> {
|
|
765
|
+
Ok(())
|
|
766
|
+
}
|
|
767
|
+
fn priority(&self) -> i32 {
|
|
768
|
+
100
|
|
769
|
+
}
|
|
770
|
+
}
|
|
771
|
+
|
|
772
|
+
let medium = Arc::new(MediumPriorityValidator);
|
|
773
|
+
let low = Arc::new(LowPriorityValidator);
|
|
774
|
+
let high_priority = Arc::new(HighPriorityValidator);
|
|
775
|
+
|
|
776
|
+
registry.register(medium).unwrap();
|
|
777
|
+
registry.register(low).unwrap();
|
|
778
|
+
registry.register(high_priority).unwrap();
|
|
779
|
+
|
|
780
|
+
let validators = registry.get_all();
|
|
781
|
+
assert_eq!(validators.len(), 3);
|
|
782
|
+
assert_eq!(validators[0].name(), "high-priority");
|
|
783
|
+
assert_eq!(validators[1].name(), "medium-priority");
|
|
784
|
+
assert_eq!(validators[2].name(), "low-priority");
|
|
785
|
+
}
|
|
786
|
+
|
|
787
|
+
#[test]
|
|
788
|
+
fn test_validator_registration_failure() {
|
|
789
|
+
struct FailingInitValidator;
|
|
790
|
+
|
|
791
|
+
impl Plugin for FailingInitValidator {
|
|
792
|
+
fn name(&self) -> &str {
|
|
793
|
+
"failing"
|
|
794
|
+
}
|
|
795
|
+
fn version(&self) -> String {
|
|
796
|
+
"1.0.0".to_string()
|
|
797
|
+
}
|
|
798
|
+
fn initialize(&self) -> Result<()> {
|
|
799
|
+
Err(KreuzbergError::Plugin {
|
|
800
|
+
message: "Init failed".to_string(),
|
|
801
|
+
plugin_name: "failing".to_string(),
|
|
802
|
+
})
|
|
803
|
+
}
|
|
804
|
+
fn shutdown(&self) -> Result<()> {
|
|
805
|
+
Ok(())
|
|
806
|
+
}
|
|
807
|
+
}
|
|
808
|
+
|
|
809
|
+
#[async_trait]
|
|
810
|
+
impl Validator for FailingInitValidator {
|
|
811
|
+
async fn validate(&self, _: &ExtractionResult, _: &ExtractionConfig) -> Result<()> {
|
|
812
|
+
Ok(())
|
|
813
|
+
}
|
|
814
|
+
fn priority(&self) -> i32 {
|
|
815
|
+
50
|
|
816
|
+
}
|
|
817
|
+
}
|
|
818
|
+
|
|
819
|
+
let mut registry = ValidatorRegistry::new();
|
|
820
|
+
let validator = Arc::new(FailingInitValidator);
|
|
821
|
+
|
|
822
|
+
let result = registry.register(validator);
|
|
823
|
+
assert!(matches!(result, Err(KreuzbergError::Plugin { .. })));
|
|
824
|
+
}
|
|
825
|
+
|
|
826
|
+
#[test]
|
|
827
|
+
fn test_validator_empty_registry() {
|
|
828
|
+
let registry = ValidatorRegistry::new();
|
|
829
|
+
let validators = registry.get_all();
|
|
830
|
+
assert_eq!(validators.len(), 0);
|
|
831
|
+
}
|
|
832
|
+
|
|
833
|
+
#[test]
|
|
834
|
+
fn test_validator_remove_and_reregister() {
|
|
835
|
+
let mut registry = ValidatorRegistry::new();
|
|
836
|
+
|
|
837
|
+
let validator: Arc<dyn Validator> = Arc::new(StrictValidator {
|
|
838
|
+
name: "validator".to_string(),
|
|
839
|
+
min_length: 5,
|
|
840
|
+
});
|
|
841
|
+
|
|
842
|
+
registry.register(Arc::clone(&validator)).unwrap();
|
|
843
|
+
assert_eq!(registry.get_all().len(), 1);
|
|
844
|
+
|
|
845
|
+
registry.remove("validator").unwrap();
|
|
846
|
+
assert_eq!(registry.get_all().len(), 0);
|
|
847
|
+
|
|
848
|
+
registry.register(validator).unwrap();
|
|
849
|
+
assert_eq!(registry.get_all().len(), 1);
|
|
850
|
+
}
|
|
851
|
+
|
|
852
|
+
#[test]
|
|
853
|
+
fn test_multiple_registries_independence() {
|
|
854
|
+
let ocr_registry = OcrBackendRegistry::new_empty();
|
|
855
|
+
let mut extractor_registry = DocumentExtractorRegistry::new();
|
|
856
|
+
let mut processor_registry = PostProcessorRegistry::new();
|
|
857
|
+
let mut validator_registry = ValidatorRegistry::new();
|
|
858
|
+
|
|
859
|
+
let extractor = Arc::new(FailingExtractor {
|
|
860
|
+
name: "test-extractor".to_string(),
|
|
861
|
+
should_fail_init: false,
|
|
862
|
+
should_fail_extract: false,
|
|
863
|
+
});
|
|
864
|
+
|
|
865
|
+
let processor = Arc::new(MetadataModifyingProcessor {
|
|
866
|
+
name: "test-processor".to_string(),
|
|
867
|
+
stage: ProcessingStage::Early,
|
|
868
|
+
});
|
|
869
|
+
|
|
870
|
+
let validator = Arc::new(StrictValidator {
|
|
871
|
+
name: "test-validator".to_string(),
|
|
872
|
+
min_length: 5,
|
|
873
|
+
});
|
|
874
|
+
|
|
875
|
+
extractor_registry.register(extractor).unwrap();
|
|
876
|
+
processor_registry.register(processor, 50).unwrap();
|
|
877
|
+
validator_registry.register(validator).unwrap();
|
|
878
|
+
|
|
879
|
+
assert_eq!(ocr_registry.list().len(), 0);
|
|
880
|
+
assert_eq!(extractor_registry.list().len(), 1);
|
|
881
|
+
assert_eq!(processor_registry.list().len(), 1);
|
|
882
|
+
assert_eq!(validator_registry.get_all().len(), 1);
|
|
883
|
+
}
|
|
884
|
+
|
|
885
|
+
#[test]
|
|
886
|
+
fn test_shutdown_all_registries() {
|
|
887
|
+
let mut ocr_registry = OcrBackendRegistry::new_empty();
|
|
888
|
+
let mut extractor_registry = DocumentExtractorRegistry::new();
|
|
889
|
+
let mut processor_registry = PostProcessorRegistry::new();
|
|
890
|
+
let mut validator_registry = ValidatorRegistry::new();
|
|
891
|
+
|
|
892
|
+
let extractor = Arc::new(FailingExtractor {
|
|
893
|
+
name: "test-extractor".to_string(),
|
|
894
|
+
should_fail_init: false,
|
|
895
|
+
should_fail_extract: false,
|
|
896
|
+
});
|
|
897
|
+
|
|
898
|
+
let processor = Arc::new(MetadataModifyingProcessor {
|
|
899
|
+
name: "test-processor".to_string(),
|
|
900
|
+
stage: ProcessingStage::Early,
|
|
901
|
+
});
|
|
902
|
+
|
|
903
|
+
let validator = Arc::new(StrictValidator {
|
|
904
|
+
name: "test-validator".to_string(),
|
|
905
|
+
min_length: 5,
|
|
906
|
+
});
|
|
907
|
+
|
|
908
|
+
extractor_registry.register(extractor).unwrap();
|
|
909
|
+
processor_registry.register(processor, 50).unwrap();
|
|
910
|
+
validator_registry.register(validator).unwrap();
|
|
911
|
+
|
|
912
|
+
ocr_registry.shutdown_all().unwrap();
|
|
913
|
+
extractor_registry.shutdown_all().unwrap();
|
|
914
|
+
processor_registry.shutdown_all().unwrap();
|
|
915
|
+
validator_registry.shutdown_all().unwrap();
|
|
916
|
+
|
|
917
|
+
assert_eq!(ocr_registry.list().len(), 0);
|
|
918
|
+
assert_eq!(extractor_registry.list().len(), 0);
|
|
919
|
+
assert_eq!(processor_registry.list().len(), 0);
|
|
920
|
+
assert_eq!(validator_registry.get_all().len(), 0);
|
|
921
|
+
}
|