Medios-Macina/docs/provider_authoring.md

# Plugin authoring: ResultTable & plugin adapters

This short guide explains how to write plugins that integrate with the *strict* ResultTable API: adapters must yield `ResultModel` instances and plugins register via `SYS.result_table_adapters.register_plugin` with a column specification and a `selection_fn`.

---

## Quick summary

- Plugins register a *plugin adapter* (callable that yields `ResultModel`).
- Plugins must also provide `columns` (static list or factory) and a `selection_fn` that returns CLI args for a selected row.
- For simple HTML table/list scraping, prefer `TableProviderMixin` from `SYS.provider_helpers` to fetch and extract rows using `SYS.html_table.extract_records`.

## Runtime dependency policy

- Treat required runtime dependencies (e.g., **Playwright**) as mandatory: import them unconditionally and let missing dependencies fail fast at import time. Avoid adding per-call try/except import guards for required modules—these silently hide configuration errors and add bloat.
- Use guarded imports only for truly optional dependencies (e.g., `pandas` for enhanced table parsing) and provide meaningful fallbacks or helpful error messages in those cases.
- Keep provider code minimal and explicit: fail early and document required runtime dependencies in README/installation notes.

---

## Minimal provider template (copy/paste)

```py
# plugins/my_plugin.py
from typing import Any, Dict, Iterable, List

from SYS.result_table_api import ResultModel, ColumnSpec, title_column, metadata_column
from SYS.result_table_adapters import register_plugin

# Example adapter: convert provider-specific items into ResultModel instances
SAMPLE_ITEMS = [
    {"name": "Example File.pdf", "path": "https://example.com/x.pdf", "ext": "pdf", "size": 1024, "source": "myprovider"},
]

def adapter(items: Iterable[Dict[str, Any]]) -> Iterable[ResultModel]:
    for it in items:
        title = it.get("name") or it.get("title") or str(it.get("path") or "")
        yield ResultModel(
            title=str(title),
            path=str(it.get("path")) if it.get("path") else None,
            ext=str(it.get("ext")) if it.get("ext") else None,
            size_bytes=int(it.get("size")) if it.get("size") is not None else None,
            metadata=dict(it),
            source=str(it.get("source")) if it.get("source") else "myprovider",
        )

# Optional: build columns dynamically from sample rows
def columns_factory(rows: List[ResultModel]) -> List[ColumnSpec]:
    cols = [title_column()]
    # add extra columns if metadata keys exist
    if any((r.metadata or {}).get("size") for r in rows):
        cols.append(ColumnSpec("size", "Size", lambda r: r.size_bytes or ""))
    return cols

# Selection args for `@N` expansion or `select` cmdlet
def selection_fn(row: ResultModel) -> List[str]:
    # prefer -path when available
    if row.path:
        return ["-path", row.path]
    return ["-title", row.title or ""]

# Register plugin (done at import time)
register_plugin("myprovider", adapter, columns=columns_factory, selection_fn=selection_fn)
```

---

## Table scraping: using TableProviderMixin (HTML tables / list-results)

If your provider scrapes HTML tables or list-like results (common on web search pages), use `TableProviderMixin`:

```py
from ProviderCore.base import Provider
from SYS.provider_helpers import TableProviderMixin

class MyTableProvider(TableProviderMixin, Provider):
    URL = ("https://example.org/search",)

    def validate(self) -> bool:
        return True

    def search(self, query: str, limit: int = 50, **kwargs):
        url = f"{self.URL[0]}?q={quote_plus(query)}"
        return self.search_table_from_url(url, limit=limit)
```

`TableProviderMixin.search_table_from_url` returns `ProviderCore.base.SearchResult` entries. If you want to integrate this plugin with the strict `ResultTable` registry, add a small adapter that converts `SearchResult` -> `ResultModel` and register it using `register_plugin` (see `Provider/vimm.py` for a real example).

---

## Columns & selection

- `columns` may be a static `List[ColumnSpec]` or a factory `def cols(rows: List[ResultModel]) -> List[ColumnSpec]` that inspects sample rows.
- `selection_fn` must accept a `ResultModel` and return a `List[str]` representing CLI args (e.g., `['-path', row.path]`). These args are used by `select` and `@N` expansion.

    **Tip:** for plugins that produce downloadable file rows prefer returning explicit URL args (e.g., `['-url', row.path]`) so the selected URL is clearly identified by downstream downloaders and to avoid ambiguous parsing when plugin hints (like `-plugin`) are present.
- Ensure your `ResultModel.source` is set (either in the model or rely on the provider name set by `serialize_row`).

---

## Optional: pandas path for `<table>` extraction

`SYS.html_table.extract_records` prefers a pure-lxml path but will use `pandas.read_html` if pandas is installed and the helper detects it works for the input table. This is optional and **not required** to author a provider — document in your provider whether it requires `pandas` and add an informative error/log message when it is missing.

---

## Testing & examples

- Write `tests/test_provider_<name>.py` that imports your provider and verifies `provider.build_table(...)` produces a `ResultTable` (has `.rows` and `.columns`) and that `serialize_rows()` yields dicts with `_selection_args`, `_selection_action` when applicable, and `source`.
- When you need to guarantee a specific CLI stage sequence (e.g., `download-file -url <path> -plugin <name>`), call `table.set_row_selection_action(row_index, tokens)` so the serialized payload emits `_selection_action` and the CLI can run the row exactly as intended.
- For table providers you can test `search_table_from_url` using a local HTML fixture or by mocking `HTTPClient` to return a small sample page.
- If you rely on pandas, add a test that monkeypatches `sys.modules['pandas']` to a simple shim to validate the pandas path.

**Example test skeleton**

```py
from SYS.result_table_adapters import get_provider
from Provider import example_provider


def test_example_provider_registration():
    plugin = get_plugin("example")
    rows = list(provider.adapter(example_provider.SAMPLE_ITEMS))
    assert rows and rows[0].title
    cols = provider.get_columns(rows)
    assert any(c.name == "title" for c in cols)
    table = provider.build_table(example_provider.SAMPLE_ITEMS)
    assert table.provider == "example" and table.rows
```

---

## References & examples

- Read `Provider/example_provider.py` for a compact example of a strict adapter and dynamic columns.
- Read `Provider/vimm.py` for a table-provider that uses `TableProviderMixin` and converts `SearchResult` → `ResultModel` for registration.
- See `docs/provider_guide.md` for a broader provider development checklist.

---

If you want, I can also add a small `Provider/myprovider_template.py` file and unit tests for it — say the word and I'll add them and wire up tests. 🎯