summaryrefslogtreecommitdiff
path: root/bin/lookup-server
blob: 510c1d494d511c81938f092b5fc41089e306e2a5 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
#!/usr/bin/env python3
# vim:tabstop=4 softtabstop=4 shiftwidth=4 textwidth=160 smarttab expandtab colorcolumn=160
#
# Copyright (C) 2021 Daniel Friesel
#
# SPDX-License-Identifier: AGPL-3.0-only

import argparse
import psycopg2
import aiohttp
from aiohttp import web
from datetime import datetime, timedelta
import dateutil.parser
from geopy.distance import distance
import json
import logging
import os
import pytz

headers = {
    "Access-Control-Allow-Origin": "*",
    "Content-Type": "application/json; charset=utf-8",
}

conn = psycopg2.connect(
    dbname=os.getenv("GEOLOOKUP_DBNAME", "geo_to_stations"),
    user=os.getenv("GEOLOOKUP_DBUSER", "geo_to_stations"),
    password=os.getenv("GEOLOOKUP_DBPASS"),
    host=os.getenv("GEOLOOKUP_DBHOST", "localhost"),
)

db_rest_api = os.getenv("GEOLOOKUP_DB_REST_API", "https://v5.db.transport.rest")

conn.autocommit = True
conn.set_session(readonly=True)

arrivals_request_count = 0
polyline_request_count = 0


def set_coarse_location(train, latlon):
    now = datetime.now(pytz.utc)
    train_evas = None
    stopovers = train["previousStopovers"]

    # includes train["stop"] -- but with arrival instead of departure
    for i, stopover in enumerate(stopovers):
        ts = None
        if stopover["departure"]:
            try:
                stopover["departure"] = dateutil.parser.parse(stopover["departure"])
                ts = stopover["departure"]
            except TypeError:
                return
        if stopover["arrival"]:
            try:
                stopover["arrival"] = dateutil.parser.parse(stopover["arrival"])
                ts = stopover["arrival"]
            except TypeError:
                return

        # start with origin. (planned)arrival is always null in a previousStopovers list except for the last entry
        # (which is the stop where arrivals were requested)
        if i > 0 and ts and ts > now:
            train_evas = (
                int(stopovers[i - 1]["stop"]["id"]),
                int(stopover["stop"]["id"]),
            )
            train_stops = (stopovers[i - 1]["stop"]["name"], stopover["stop"]["name"])
            train_coords = (
                (
                    stopovers[i - 1]["stop"]["location"]["latitude"],
                    stopovers[i - 1]["stop"]["location"]["longitude"],
                ),
                (
                    stopover["stop"]["location"]["latitude"],
                    stopover["stop"]["location"]["longitude"],
                ),
            )
            # XXX known bug: we're saving departure at i-1 and (possibly) departure at i. For a more accurate coarse position estimate later on,
            # we need to track departure at i-1 and arrival at i. But we don't always have it.
            train_times = (stopovers[i - 1]["departure"], ts)
            break
    if not train_evas:
        return

    if not train_times[0]:
        return

    train["evas"] = train_evas
    train["stop_names"] = train_stops
    train["coords"] = train_coords
    train["times"] = train_times

    train["progress_ratio"] = 1 - (
        (train["times"][1].timestamp() - now.timestamp())
        / (train["times"][1].timestamp() - train["times"][0].timestamp())
    )
    train["progress_ratio"] = max(0, min(1, train["progress_ratio"]))

    if train["progress_ratio"] == 0:
        train["location"] = train["coarse_location"] = train["coords"][0]
    elif train["progress_ratio"] == 1:
        train["location"] = train["coarse_location"] = train["coords"][1]
    else:
        ratio = train["progress_ratio"]
        coords = train["coords"]
        train["coarse_location"] = (
            coords[1][0] * ratio + coords[0][0] * (1 - ratio),
            coords[1][1] * ratio + coords[0][1] * (1 - ratio),
        )

    if distance(train["coords"][0], train["coords"][1]).km < 20:
        # do not request polyline if the train is between stops less than 20km apart. This speeds up requests
        # (and reduces transport.rest load) at a hopefully low impact on accuracy.
        train["location"] = train["coarse_location"]

    if train_evas[1] == int(train["stop"]["id"]):
        # we can compare departure at previous stop with arrival at this stop. this is most accurate for position estimation.
        train["preferred"] = True
    else:
        train["preferred"] = False

    # the time (i.e., number of minutes) the train needs to travel to reach the requested position
    # might be a better metric than raw distance.
    train["distance"] = distance(train["coarse_location"], latlon).km


async def set_location(train):
    trip_id = train["tripId"]
    line = train["line"]["name"]
    url = f"{db_rest_api}/trips/{trip_id}?lineName={line}&polyline=true"
    return
    logging.debug(f"Requesting polyline for {line}: {url}")
    global polyline_request_count
    polyline_request_count += 1
    async with aiohttp.ClientSession() as session:
        async with session.get(url) as response:
            content = await response.text()
            content = json.loads(content)


def is_in_transit(train):
    return 0 < train["progress_ratio"] < 1


def format_train(train):
    train_type, line_no = train["line"]["name"].split()
    train_no = train["line"]["fahrtNr"]
    return {
        "line": f"{train_type} {line_no}",
        "train": f"{train_type} {train_no}",
        "tripId": train["tripId"],
        "location": train["coarse_location"],
        "distance": round(train["distance"], 1),
        "stops": [
            (
                train["evas"][0],
                train["stop_names"][0],
                train["times"][0].strftime("%H:%M"),
            ),
            (
                train["evas"][1],
                train["stop_names"][1],
                train["times"][1].strftime("%H:%M"),
            ),
        ],
    }


async def handle_stats(request):
    response = {
        "arrivals_request_count": arrivals_request_count,
        "polyline_request_count": polyline_request_count,
    }
    return web.Response(body=json.dumps(response), headers=headers)


async def handle_search(request):
    try:
        lat = float(request.query.get("lat"))
        lon = float(request.query.get("lon"))
    except TypeError:
        return web.HTTPBadRequest(text="lat/lon are mandatory")
    except ValueError:
        return web.HTTPBadRequest(text="lat/lon must be floating-point numbers")

    lut_lat = round(lat * 1000)
    lut_lon = round(lon * 1000)

    evas = set()

    with conn.cursor() as cur:
        cur.execute(
            "select stations from stations where lat between %s and %s and lon between %s and %s",
            (lut_lat - 3, lut_lat + 3, lut_lon - 3, lut_lon + 3),
        )
        for eva_list in cur.fetchall():
            evas.update(eva_list[0])

    if not evas:
        response = {"evas": list(), "trains": list()}
        return web.Response(body=json.dumps(response), headers=headers)

    arrivals = list()
    trains = list()

    # deliberately not parallelized to minimize load on transport.rest
    for eva in evas:
        logging.debug(f"Requesting arrivals at {eva}")
        global arrivals_request_count
        arrivals_request_count += 1
        async with aiohttp.ClientSession() as session:
            async with session.get(
                f"{db_rest_api}/stops/{eva}/arrivals?results=40&duration=120&stopovers=true&bus=false&subway=false&tram=false"
            ) as response:
                content = await response.text()
                content = json.loads(content)
                arrivals.append(content)

    for train_list in arrivals:
        for train in train_list:
            is_candidate = False
            for stop in train["previousStopovers"]:
                if (
                    int(stop["stop"]["id"]) in evas
                    and stop["stop"]["id"] != train["stop"]["id"]
                ):
                    is_candidate = True
                    break
            if is_candidate:
                trains.append(train)

    logging.debug(f"{len(trains)} trains travel between at least two requested evas")

    for train in trains:
        set_coarse_location(train, (lat, lon))

    trains = list(filter(lambda train: "coarse_location" in train, trains))
    logging.debug(f"{len(trains)} trains have a coarse location")

    trains = sorted(
        trains, key=lambda train: 0 if train["preferred"] else train["distance"]
    )

    # remove duplicates. for now, we keep the preferred version, or the one with the lowest estimated distance.
    # later on, we'll need to request polylines and perform accurate calculations.
    # TODO polyline requests are not needed for trains currently located at a station (ratio == 0 / == 1)
    # It should also be fine to skip them if the distance between stops[0] and stops[1] is less than ~ 20km
    # Wenn sich ein Zug gerade an einem Bahnhof befindet (ratio == 0 / == 1) und mehrere km entfernt ist kann man ihn auch direkt ganz rausfiltern
    seen = set()
    trains = [
        seen.add(train["line"]["fahrtNr"]) or train
        for train in trains
        if train["line"]["fahrtNr"] not in seen
    ]

    logging.debug(f"{len(trains)} trains remain after deduplication")
    need_fine = list(filter(lambda train: "location" not in train, trains))
    need_fine = list(filter(is_in_transit, trains))
    logging.debug(f"{len(need_fine)} trains need a polyline")

    for train in trains:
        await set_location(train)

    trains = sorted(trains, key=lambda train: train["distance"])
    trains = list(map(format_train, trains[:10]))

    response = {"evas": list(evas), "trains": trains}

    return web.Response(body=json.dumps(response, ensure_ascii=False), headers=headers)


if __name__ == "__main__":

    parser = argparse.ArgumentParser(
        description="geolocation to train estimation service"
    )
    parser.add_argument(
        "--log-level",
        metavar="LEVEL",
        choices=["debug", "info", "warning", "error"],
        default="warning",
        help="Set log level",
    )
    parser.add_argument("--port", type=int, metavar="PORT", default=8080)
    parser.add_argument("--prefix", type=str, metavar="PATH", default="/")
    args = parser.parse_args()

    if args.log_level:
        numeric_level = getattr(logging, args.log_level.upper(), None)
        if not isinstance(numeric_level, int):
            print(f"Invalid log level: {args.log_level}", file=sys.stderr)
            sys.exit(1)
        logging.basicConfig(level=numeric_level)

    app = web.Application()
    app.add_routes(
        [
            web.get(f"{args.prefix}search", handle_search),
            web.get(f"{args.prefix}stats", handle_stats),
        ]
    )
    web.run_app(app, host="localhost", port=args.port)