added tests of remaining transformers

[epinzur]

Adds a new option to mark tests as extra. This requires the --runextras flag on pytest to run.

Adds tests for the transformers that were missing tests. Many require extras and are not tested against python 3.13.

Also simplifies the HierarchyLinkExtractor, now called the ParentTransformer.

This closes #83 and #59.

[github-actions]

Test Results

    8 files  ± 0    8 suites  ±0   2m 1s ⏱️ +26s
  328 tests + 5  328 ✅ + 5    0 💤 ±0  0 ❌ ±0 
1 312 runs  +20  780 ✅ +16  532 💤 +4  0 ❌ ±0 

Results for commit 583937b. ± Comparison against base commit c93576b.

This pull request removes 2 and adds 7 tests. Note that renamed tests count towards both.

packages.langchain-graph-retriever.tests.document_transformers.test_metadata_denormalizer ‑ test_revert_documents
packages.langchain-graph-retriever.tests.document_transformers.test_metadata_denormalizer ‑ test_transform_documents

packages.langchain-graph-retriever.tests.transformers.test_gliner ‑ test_transform_documents
packages.langchain-graph-retriever.tests.transformers.test_hierarchy ‑ test_transform_documents
packages.langchain-graph-retriever.tests.transformers.test_html_hyperlink ‑ test_transform_documents
packages.langchain-graph-retriever.tests.transformers.test_keybert ‑ test_transform_documents
packages.langchain-graph-retriever.tests.transformers.test_metadata_denormalizer ‑ test_revert_documents
packages.langchain-graph-retriever.tests.transformers.test_metadata_denormalizer ‑ test_transform_documents
packages.langchain-graph-retriever.tests.transformers.test_spacy ‑ test_transform_documents

♻️ This comment has been updated with latest results.

[bjchambers]

Nice. I think it's useful to have this transformer (especially for LangFlow, etc.) but this could be a good thing to have an example of doing too. Specifically, if the user has already written code to put the path in the metadata... it wouldn't be too hard to just write the parent themselves.

We should definitely have some examples showing "here is how to set the metadata for X" rather than relying on hiding it all in transformers. Specifically, that enables the user to encode their own structured metadata.

This would be a good example to use, since it demonstrates "setting something like parent in the metadata enables different kinds of navigation via the edges".

[coveralls]

Pull Request Test Coverage Report for Build 13164213708

Details

• 43 of 47 (91.49%) changed or added relevant lines in 5 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+10.3%) to 92.916%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
packages/langchain-graph-retriever/src/langchain_graph_retriever/transformers/keybert.py	6	7	85.71%
packages/langchain-graph-retriever/src/langchain_graph_retriever/transformers/html_hyperlink.py	9	12	75.0%

Totals
Change from base Build 13163802757:	10.3%
Covered Lines:	1468
Relevant Lines:	1554

---

💛 - Coveralls