Refactor DepMapData.DataNamespace into reusable CancerDataNamespace, add custom dataset injection, and exclude non-core datasets from built-in defaults

[Copilot]

DepMapData.DataNamespace was structurally minimal and only supported read access to pre-defined datasets. This update modernizes namespace handling, introduces a reusable parent namespace class for future database integrations, adds explicit APIs for users to attach new in-memory datasets directly to a DepMapData instance, and narrows built-in DepMap defaults to core datasets only.

• Reusable namespace base for broader data sources

  • Extracted shared namespace behavior into candi/data/cancer_database.py as CancerDataNamespace.
  • Updated DepMapData.DataNamespace to inherit from CancerDataNamespace, preserving DepMap-specific typed dataset annotations while making the core behavior reusable for other database handlers.

• Namespace structure and ergonomics

  • Refactored namespace implementation with stronger class shape (__slots__, typed access behavior, clearer docstring).
  • Added __dir__ support so interactive discovery includes both file-backed and runtime-added datasets.

• Custom dataset injection

  • Added DataNamespace.add(name, dataset, overwrite=False) with input validation:
    • non-empty string names only
    • valid Python identifier requirement (for attribute-style access)
    • overwrite protection unless explicitly enabled
  • Added attribute-assignment path for custom datasets via __setattr__ (except internal _parent).
  • Added DepMapData.add_dataset(name, dataset, overwrite=False) as a top-level convenience API.

• Availability semantics

  • Updated list_available() to return both predefined dataset names and user-added runtime datasets.

• Built-in dataset scope correction

  • Removed OmicsProteinAbundance and CRISPRScreenMap from built-in/default DepMap dataset definitions.
  • Removed CRISPRScreenMap.csv from DepMapAPI.download_essential().
  • These datasets are now expected to be added manually via the new custom dataset APIs when needed.

depmap = DepMapData(data_dir="...")

# explicit API
depmap.data.add("MyScores", my_df)

# convenience API on parent
depmap.add_dataset("MyAnnData", my_anndata)

# attribute-style injection
depmap.data.CustomPanel = panel_df

# attribute-style retrieval
depmap.data.MyScores

[abearab]

@copilot There are two datasets I need you to exclude from the main code. Since these are not part of main DepMap downloader API so I think it's best if users can add them manually using new functionalities introduced in this PR

1. DepMap proteomics data OmicsProteinAbundance: pd.DataFrame
2. PRISM drug sensitivity data CRISPRScreenMap: ad.AnnData

[Copilot]

@copilot There are two datasets I need you to exclude from the main code. Since these are not part of main DepMap downloader API so I think it's best if users can add them manually using new functiona...

Implemented in commit 3edc028. I removed OmicsProteinAbundance and CRISPRScreenMap from the built-in defaults (namespace hints, default dataset paths, and essential downloader list), so they can now be supplied manually through the new add-dataset functionality.

[abearab]

@copilot make the DataNamespace a parental class for depmap class so it can be used for other type of databases. Maybe name it cancer database or something general so I can expand to other data sources. It's better if it's located in a different python file in data submodule in candi

[Copilot]

@copilot make the DataNamespace a parental class for depmap class so it can be used for other type of databases. Maybe name it cancer database or something general so I can expand to other data sour...

Done in commit 118157b. I extracted a reusable parent namespace into candi/data/cancer_database.py as CancerDataNamespace and updated DepMapData.DataNamespace to inherit from it, so the same namespace base can be reused for other database handlers.

[abearab]

@copilot

[Copilot]

Pull request overview

This PR refactors DepMap dataset namespace handling to make .data attribute access reusable across cancer data backends, adds APIs to inject in-memory datasets at runtime, and narrows the built-in DepMap default dataset set to core datasets only.

Changes:

• Introduces a reusable CancerDataNamespace base class (attribute access, add(), __dir__) and makes DepMapData.DataNamespace inherit from it.
• Adds DepMapData.add_dataset() and expands list_available() to include runtime-injected datasets.
• Removes CRISPRScreenMap/OmicsProteinAbundance from built-in DepMap defaults and from “essential” downloads/paths/loading.

Reviewed changes

Copilot reviewed 2 out of 3 changed files in this pull request and generated 4 comments.

File	Description
candi/data/depmap.py	Refactors DataNamespace to inherit from shared base, adds dataset injection/listing, and removes non-core built-in datasets from defaults/downloads.
candi/data/_database.py	Adds CancerDataNamespace base implementation for attribute-style dataset access + runtime injection.
candi/data/__init__.py	Minor formatting/cleanup of exports line.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[abearab]

@copilot