backend/app/api/data.py

"""Data import / sample dataset endpoints."""

from __future__ import annotations

import csv
import io
import json
from pathlib import Path

from fastapi import APIRouter, File, HTTPException, UploadFile

from app.api.schemas import SampleDatasetModel, SampleInfoModel

SAMPLES_DIR = Path(__file__).resolve().parents[1] / "data_sources" / "samples"

router = APIRouter(prefix="/api/data", tags=["data"])

SAMPLE_REGISTRY = {
    "gdp": {"file": "sample-gdp.json", "icon": "📊"},
    "climate": {"file": "sample-climate.json", "icon": "🌡️"},
    "population": {"file": "sample-population.json", "icon": "👥"},
}


def _load(name: str) -> dict:
    path = SAMPLES_DIR / SAMPLE_REGISTRY[name]["file"]
    if not path.exists():
        raise HTTPException(status_code=404, detail=f"sample {name} not found")
    return json.loads(path.read_text(encoding="utf-8"))


@router.get("/samples", response_model=list[SampleInfoModel])
async def list_samples() -> list[SampleInfoModel]:
    out: list[SampleInfoModel] = []
    for sid, meta in SAMPLE_REGISTRY.items():
        if sid not in SAMPLE_REGISTRY:
            continue
        data = _load(sid)
        out.append(
            SampleInfoModel(
                id=sid,
                name=data.get("name", sid),
                name_en=data.get("name_en"),
                source=data.get("source", ""),
                unit=data.get("unit", ""),
                period=data.get("period", ""),
                description=data.get("description", ""),
                n=len(data.get("values", [])),
                icon=meta["icon"],
            )
        )
    return out


@router.get("/samples/{sid}", response_model=SampleDatasetModel)
async def get_sample(sid: str) -> SampleDatasetModel:
    if sid not in SAMPLE_REGISTRY:
        raise HTTPException(status_code=404, detail=f"unknown sample id {sid}")
    data = _load(sid)
    return SampleDatasetModel(
        id=sid,
        name=data.get("name", sid),
        name_en=data.get("name_en"),
        source=data.get("source", ""),
        unit=data.get("unit", ""),
        period=data.get("period", ""),
        description=data.get("description", ""),
        n=len(data.get("values", [])),
        icon=SAMPLE_REGISTRY[sid]["icon"],
        values=list(data.get("values", [])),
    )


@router.post("/upload")
async def upload_csv(file: UploadFile = File(...)) -> dict:
    """Accept a CSV upload, parse the first numeric column, return values + metadata.

    Designed for the data-import workflow; the frontend renders a preview and
    feeds the values back into /bayesian/compute.
    """
    if file.content_type and "csv" not in file.content_type and not file.filename.endswith(".csv"):
        raise HTTPException(status_code=415, detail="please upload a CSV file")
    raw = await file.read()
    try:
        text = raw.decode("utf-8-sig")
    except UnicodeDecodeError:
        text = raw.decode("latin-1")

    reader = csv.reader(io.StringIO(text))
    rows = list(reader)
    if not rows:
        raise HTTPException(status_code=400, detail="empty CSV")

    # Detect header
    header = rows[0]
    has_header = any(not _is_number(c) for c in header)
    data_rows = rows[1:] if has_header else rows

    # Find first column with at least 50% numeric values
    n_cols = max((len(r) for r in data_rows), default=0)
    chosen_col = 0
    best_score = -1.0
    for col in range(n_cols):
        nums = sum(1 for r in data_rows if col < len(r) and _is_number(r[col]))
        score = nums / max(len(data_rows), 1)
        if score > best_score:
            best_score = score
            chosen_col = col

    values: list[float] = []
    skipped = 0
    for r in data_rows:
        if chosen_col >= len(r):
            skipped += 1
            continue
        cell = r[chosen_col].strip()
        if not cell:
            skipped += 1
            continue
        try:
            values.append(float(cell))
        except ValueError:
            skipped += 1

    if len(values) < 2:
        raise HTTPException(
            status_code=400,
            detail=f"could not parse at least 2 numeric values (got {len(values)})",
        )

    return {
        "filename": file.filename,
        "values": values,
        "n_total_rows": len(data_rows),
        "n_used": len(values),
        "n_skipped": skipped,
        "chosen_column_index": chosen_col,
        "chosen_column_label": header[chosen_col]
        if has_header and chosen_col < len(header)
        else f"column_{chosen_col}",
    }


def _is_number(s: str) -> bool:
    try:
        float(s.strip())
        return True
    except (ValueError, AttributeError):
        return False