"""c_artifact_get (`GET /api/artifacts/:artifact_id`, named in
docs/sdd/sdd-mvcp-contract.json but never documented in 03-C's prose, and
never implemented) + c_artifact_list (no operation_id in either the prose
doc or the MVCP JSON at all, but v_artifact_browser's own filter fields --
`artifact_type_filter`/`task_id_filter`, 02-V §2.8 -- require one; same
"unowned gap" precedent as c_task_list before U-025 built it). Returns a
bare array from list_artifacts(), matching list_tasks()'s shape rather than
the MVCP JSON's c_task_list `{items, total}` wrapper -- no pagination need
for a local single-project MVP, and staying consistent with this
codebase's own established list-endpoint shape matters more here than
matching a machine-generated mirror of the SDD.
"""

import json
import os


class ArtifactNotFoundError(Exception):
    """NOT_FOUND -- c_artifact_get on a nonexistent artifact_id."""


def _row_to_dict(r):
    return {
        "artifact_id": r[0], "artifact_type": r[1], "task_id": r[2], "produced_by_run_id": r[3],
        "file_path": r[4], "git_commit_sha": r[5],
        "depends_on_artifact_ids": json.loads(r[6]) if r[6] else [],
        "validation_status": r[7], "created_at": r[8],
    }


def list_artifacts(conn, project_id, task_id=None, artifact_type=None):
    query = (
        "SELECT a.artifact_id, a.artifact_type, a.task_id, a.produced_by_run_id, a.file_path, "
        "a.git_commit_sha, a.depends_on_artifact_ids, a.validation_status, a.created_at "
        "FROM artifact a JOIN task t ON t.task_id = a.task_id WHERE t.project_id = ?"
    )
    params = [project_id]
    if task_id:
        query += " AND a.task_id = ?"
        params.append(task_id)
    if artifact_type:
        query += " AND a.artifact_type = ?"
        params.append(artifact_type)
    query += " ORDER BY a.created_at DESC"
    rows = conn.execute(query, params).fetchall()
    return [_row_to_dict(r) for r in rows]


def get_artifact(conn, artifact_id):
    """Returns the artifact row plus its file content (read live off disk --
    `artifact` itself has no content column, see 01-M 1.5) and its lineage
    chain, so v_artifact_browser's preview pane needs exactly one request."""
    row = conn.execute(
        "SELECT a.artifact_id, a.artifact_type, a.task_id, a.produced_by_run_id, a.file_path, "
        "a.git_commit_sha, a.depends_on_artifact_ids, a.validation_status, a.created_at, p.target_folder_path "
        "FROM artifact a JOIN task t ON t.task_id = a.task_id JOIN project p ON p.project_id = t.project_id "
        "WHERE a.artifact_id = ?",
        (artifact_id,),
    ).fetchone()
    if row is None:
        raise ArtifactNotFoundError(artifact_id)

    result = _row_to_dict(row[:9])
    abs_path = os.path.join(row[9], row[4])
    try:
        with open(abs_path, "r", encoding="utf-8", errors="replace") as f:
            result["content"] = f.read()
    except OSError:
        result["content"] = None
    result["lineage"] = lineage_chain(conn, artifact_id)
    return result


def lineage_chain(conn, artifact_id, max_depth=50):
    """02-V §2.8: click_row 展示 depends_on_artifact_ids 血缘链. Returns
    oldest-first [{artifact_id, artifact_type}, ...] ending with
    `artifact_id` itself. `depends_on_artifact_ids` is a list (an artifact
    can depend on more than one predecessor), but every real usage in this
    pipeline is a linear brd->prd->sdd chain and the wireframe itself only
    ever shows a single line -- walks depends_on_artifact_ids[0] when more
    than one dependency exists, a deliberate simplification for a case this
    codebase doesn't otherwise produce. A cycle guard is defensive: this is
    a read path and shouldn't assume c_artifact_write's own dependency
    validation was never bypassed (e.g. by a direct DB edit)."""
    chain = []
    seen = set()
    current_id = artifact_id
    while current_id and current_id not in seen and len(chain) < max_depth:
        seen.add(current_id)
        row = conn.execute(
            "SELECT artifact_id, artifact_type, depends_on_artifact_ids FROM artifact WHERE artifact_id = ?",
            (current_id,),
        ).fetchone()
        if row is None:
            break
        chain.append({"artifact_id": row[0], "artifact_type": row[1]})
        deps = json.loads(row[2]) if row[2] else []
        current_id = deps[0] if deps else None
    chain.reverse()
    return chain