import time

import prometheus_client
from prometheus_client.registry import Collector
from prometheus_client.core import GaugeMetricFamily, REGISTRY

from tilekiln.storage import Storage

# Disable default metrics since we're not monitoring this process, we're monitoring
# the DB sizes
REGISTRY.unregister(prometheus_client.GC_COLLECTOR)
REGISTRY.unregister(prometheus_client.PLATFORM_COLLECTOR)
REGISTRY.unregister(prometheus_client.PROCESS_COLLECTOR)

# Don't auto-create metrics for when the metric was created
prometheus_client.disable_created_metrics()


class TilekilnCollector(Collector):
    def __init__(self, storage: Storage):
        self.__storage = storage
        super().__init__()

        self.__i = 0

    # This one is run every 15s
    def collect(self):
        # This is manually producing the metrics described in
        # https://prometheus.io/docs/concepts/metric_types/#summary
        # Native histograms would be nice here, but are still only experimental
        size = GaugeMetricFamily('tilekiln_stored_bytes_sum', 'Total size of tiles',
                                 labels=['tileset', 'zoom'])
        quantiles = GaugeMetricFamily('tilekiln_stored_bytes', 'Tile percentiles',
                                      labels=['tileset', 'zoom', 'quantile'])
        total = GaugeMetricFamily('tilekiln_stored_count', 'Tiles in tilekiln storage',
                                  labels=['tileset', 'zoom'])
        for metric in self.__storage.metrics():
            size.add_metric([metric.id, str(metric.zoom)], metric.size)
            total.add_metric([metric.id, str(metric.zoom)], metric.num_tiles)
            for i in range(0, len(metric.percentiles[0])):
                quantiles.add_metric([metric.id, str(metric.zoom), str(metric.percentiles[0][i])],
                                     metric.percentiles[1][i])
        yield total
        yield size
        yield quantiles

    def update(self):
        self.__i = self.__i + 1
        pass


METRIC_UPDATE_TIME = prometheus_client.Summary('tilekiln_metrics_storage_seconds',
                                               'Time spent updating metrics')


@METRIC_UPDATE_TIME.time()
def monitored_update_metrics(storage: Storage):
    '''Update storage metrics while tracking call time

    The easiest way to monitor a function is to annotate it. Rather than require
    prometheus in storage.py, we wrap it and annotate the wrapper to track call time.
    '''
    storage.update_metrics()


def serve_prometheus(storage: Storage, addr, port, sleep):
    '''Start a prometheus server for storage info.'''
    collector = TilekilnCollector(storage)
    REGISTRY.register(collector)
    # start http
    prometheus_client.start_http_server(port=port, addr=addr)
    while True:
        # TODO: Time this with prometheus
        monitored_update_metrics(storage)
        time.sleep(sleep)