import csv
import json
import os
from typing import TYPE_CHECKING

import boto3

if TYPE_CHECKING:
    from mypy_boto3_sqs.type_defs import SendMessageBatchRequestEntryTypeDef

ENVIRONMENT = os.environ.get("ENVIRONMENT", "qa")
CSV_FILE = os.environ["CSV_FILE"]
START_OFFSET = int(os.environ.get("START_OFFSET", 0))
# Set OVERWRITE=true to re-scan assets that already have a scan: each message body
# carries an "OVERWRITE" flag that the hive_ai_detection lambda (lambda-assets) reads
# to post with overwrite, so ows-assets replaces the existing rows instead of 409ing.
# The "OVERWRITE" key name is a contract with that consumer. Fail loud on any value
# other than true/false so a typo (e.g. OVERWRITE=1) can't silently run a no-op.
_overwrite = os.environ.get("OVERWRITE", "false").lower()
if _overwrite not in {"true", "false"}:
    raise SystemExit(f"OVERWRITE must be 'true' or 'false', got {_overwrite!r}")
OVERWRITE = _overwrite == "true"

QUEUE_NAME = f"{ENVIRONMENT}-hive-ai-detection-backfill-queue"
SEND_MESSAGE_BATCH_MAXIMUM = 10
PROGRESS_INTERVAL_MESSAGES = 10_000


def main() -> None:
    sqs = boto3.client("sqs")
    queue_url = sqs.get_queue_url(QueueName=QUEUE_NAME)["QueueUrl"]

    with open(f"input/{CSV_FILE}") as csv_file:
        asset_rows = list(csv.DictReader(csv_file))

    total = len(asset_rows)
    print(
        f"{total} assets in {CSV_FILE}; starting at offset {START_OFFSET}; "
        f"overwrite={OVERWRITE}"
    )

    overwrite_field = {"OVERWRITE": True} if OVERWRITE else {}

    enqueued_message_total = 0
    for offset in range(START_OFFSET, total, SEND_MESSAGE_BATCH_MAXIMUM):
        batch = asset_rows[offset : offset + SEND_MESSAGE_BATCH_MAXIMUM]
        entries: list[SendMessageBatchRequestEntryTypeDef] = [
            {
                "Id": str(index),
                "MessageBody": json.dumps(
                    {
                        "ASSET_FINAL_ID": int(row["ID"]),
                        "DURATION_MS": int(row["DURATION"]),
                        **overwrite_field,
                    }
                ),
            }
            for index, row in enumerate(batch)
        ]
        response = sqs.send_message_batch(QueueUrl=queue_url, Entries=entries)
        if response.get("Failed"):
            # rerun with START_OFFSET=<offset> to resume; duplicates from the
            # partial batch re-scan and then 409 at the save step — bounded to
            # one batch, so the wasted cost is negligible
            raise RuntimeError(
                f"send_message_batch failed at offset {offset}: {response['Failed']}"
            )

        enqueued_message_total += len(batch)
        enqueued = offset + len(batch)
        if enqueued % PROGRESS_INTERVAL_MESSAGES < SEND_MESSAGE_BATCH_MAXIMUM:
            print(f"{enqueued}/{total} enqueued")

    print(f"Done: {enqueued_message_total} messages enqueued to {QUEUE_NAME}")