Performance fixes and refactorings

[roger-mahler]

New features

• API endpoint /speeches now accepts speech ids (UUID) (#54):

class CommonQueryParams(SpeakerQueryParams):
    def __init__(...),
        ...
        speech_id: List[str] = Query(None, description="List of speech ids to filter by"),

How to use in call:

    speech_ids: list[str] = [ ... ]
    url: str = f"{version}/tools/speeches"
    json: dict = {'speech_id': speech_ids}
    response: Response = fastapi_client.post(url, json=json)

• API endpoint /speeches now accepts POST requests (#54)

• Updates of values in returned speech index is now configurable in config.yaml (https://github.com/humlab-swedeb/swedeb-api/commit/71364c1dd3a0f9e36cdc9e63e2f2b6a3ad74c9cd).

Breaking Changes:

• Schema for /speeches endpoint has changed. Bold attributes in the following list are new attributes, and attribute formatted_speech_id has changed name to speech_name.

    name: str = Field(None, description="Name of speaker")
    year: int = Field(None, description="Year of speech", examples=[1960])
    gender: str = Field(None, description="Gender of speaker")
    **gender_abbrev**: str = Field(None, description="Gender of speaker")
    party_abbrev: str = Field(None, description="Party of speaker")
    speech_link: str = Field(None, description="Source of speech")
    document_name: str = Field(None, description="Unique id of speech")
    link: str = Field(None, description="Link to the speaker")
    **speech_name**: str = Field(None, description="Formatted speech id")
    **chamber_abbrev**: str = Field(None, description="Chamber of speech")
    **speech_id**: str = Field(None, description="Unique id of speech")
    **wiki_id**: str = Field(None, description="Wiki id of speaker")
    **document_id**: int = Field(None, description="Document system id")

Refactorings and changes

• Consolidated logic for fetching speeches (https://github.com/humlab-swedeb/swedeb-api/commit/777fe1e06ffe9879755f3299a488ef9363cdf6e9)
• API for fetching speeches now accepts either speech id (UUID) or document name (https://github.com/humlab-swedeb/swedeb-api/blob/720e1cf14c61a4b4de5453188f3f6ab541772c0b/api_swedeb/core/speech_text.py)
• Removed from_year and to_year parameters (added to filter opts instead). (https://github.com/humlab-swedeb/swedeb-api/commit/337b0e034c8916663006c9a5124084fd79f21af8, https://github.com/humlab-swedeb/swedeb-api/commit/18dad3f79b32d77447c3a328b4d5111dc1cf337e, https://github.com/humlab-swedeb/swedeb-api/commit/9f08fce8e551703d70548e96059ebdcd67badeb9, https://github.com/humlab-swedeb/swedeb-api/commit/c499bdfba7858d5965e4d4e32060bc30cc57783f)
• Speaker can now be retrieved using person_id, wiki_id or internal system id. (https://github.com/humlab-swedeb/swedeb-api/commit/7f5a4a422f0288cbc53bd8683ceee8f8fa11c207)

Performance fixes

• Faster retrieval of speeches, code no longer use DTM corpus, instead mask-operation on document index.
• Retrieval of speech text has optimized.
• Consolidated and updated code for decoding speech index (https://github.com/humlab-swedeb/swedeb-api/commit/ad6811ea20fb97060934ea720655dcf3574e0391, https://github.com/humlab-swedeb/swedeb-api/commit/daccae9c0b59499357d79d2abdb326664d9cdbd2)
• Several other perfomance fixes.

Bug fixes

• Removed use of fill_gaps when computing trends (https://github.com/humlab-swedeb/swedeb-api/commit/777fe1e06ffe9879755f3299a488ef9363cdf6e9).
• Several other fixes.