]> git.openstreetmap.org Git - nominatim.git/blobdiff - nominatim/indexer/indexer.py
don't even try heavily penalized searches
[nominatim.git] / nominatim / indexer / indexer.py
index 094d127987bb5d7c6b88f007d83c2e1481aba9e2..233423f03c6a202ec088cfeb0fe7ac26c79db01f 100644 (file)
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# This file is part of Nominatim. (https://nominatim.org)
+#
+# Copyright (C) 2022 by the Nominatim developer community.
+# For a full list of authors see the git log.
 """
 Main work horse for indexing (computing addresses) the database.
 """
-# pylint: disable=C0111
+from typing import Optional, Any, cast
 import logging
-import select
+import time
 
-import psycopg2
+import psycopg2.extras
 
-from .progress import ProgressLogger
-from ..db.async_connection import DBConnection
+from nominatim.tokenizer.base import AbstractTokenizer
+from nominatim.indexer.progress import ProgressLogger
+from nominatim.indexer import runners
+from nominatim.db.async_connection import DBConnection, WorkerPool
+from nominatim.db.connection import connect, Connection, Cursor
+from nominatim.typing import DictCursorResults
 
 LOG = logging.getLogger()
 
-class RankRunner:
-    """ Returns SQL commands for indexing one rank within the placex table.
+
+class PlaceFetcher:
+    """ Asynchronous connection that fetches place details for processing.
     """
+    def __init__(self, dsn: str, setup_conn: Connection) -> None:
+        self.wait_time = 0.0
+        self.current_ids: Optional[DictCursorResults] = None
+        self.conn: Optional[DBConnection] = DBConnection(dsn,
+                                               cursor_factory=psycopg2.extras.DictCursor)
+
+        with setup_conn.cursor() as cur:
+            # need to fetch those manually because register_hstore cannot
+            # fetch them on an asynchronous connection below.
+            hstore_oid = cur.scalar("SELECT 'hstore'::regtype::oid")
+            hstore_array_oid = cur.scalar("SELECT 'hstore[]'::regtype::oid")
+
+        psycopg2.extras.register_hstore(self.conn.conn, oid=hstore_oid,
+                                        array_oid=hstore_array_oid)
+
+    def close(self) -> None:
+        """ Close the underlying asynchronous connection.
+        """
+        if self.conn:
+            self.conn.close()
+            self.conn = None
 
-    def __init__(self, rank):
-        self.rank = rank
 
-    def name(self):
-        return "rank {}".format(self.rank)
+    def fetch_next_batch(self, cur: Cursor, runner: runners.Runner) -> bool:
+        """ Send a request for the next batch of places.
+            If details for the places are required, they will be fetched
+            asynchronously.
 
-    def sql_count_objects(self):
-        return """SELECT count(*) FROM placex
-                  WHERE rank_address = {} and indexed_status > 0
-               """.format(self.rank)
+            Returns true if there is still data available.
+        """
+        ids = cast(Optional[DictCursorResults], cur.fetchmany(100))
 
-    def sql_get_objects(self):
-        return """SELECT place_id FROM placex
-                  WHERE indexed_status > 0 and rank_address = {}
-                  ORDER BY geometry_sector""".format(self.rank)
+        if not ids:
+            self.current_ids = None
+            return False
 
-    @staticmethod
-    def sql_index_place(ids):
-        return "UPDATE placex SET indexed_status = 0 WHERE place_id IN ({})"\
-               .format(','.join((str(i) for i in ids)))
+        assert self.conn is not None
+        self.current_ids = runner.get_place_details(self.conn, ids)
 
+        return True
 
-class InterpolationRunner:
-    """ Returns SQL commands for indexing the address interpolation table
-        location_property_osmline.
-    """
+    def get_batch(self) -> DictCursorResults:
+        """ Get the next batch of data, previously requested with
+            `fetch_next_batch`.
+        """
+        assert self.conn is not None
+        assert self.conn.cursor is not None
 
-    @staticmethod
-    def name():
-        return "interpolation lines (location_property_osmline)"
-
-    @staticmethod
-    def sql_count_objects():
-        return """SELECT count(*) FROM location_property_osmline
-                  WHERE indexed_status > 0"""
-
-    @staticmethod
-    def sql_get_objects():
-        return """SELECT place_id FROM location_property_osmline
-                  WHERE indexed_status > 0
-                  ORDER BY geometry_sector"""
-
-    @staticmethod
-    def sql_index_place(ids):
-        return """UPDATE location_property_osmline
-                  SET indexed_status = 0 WHERE place_id IN ({})"""\
-               .format(','.join((str(i) for i in ids)))
-
-class BoundaryRunner:
-    """ Returns SQL commands for indexing the administrative boundaries
-        of a certain rank.
-    """
+        if self.current_ids is not None and not self.current_ids:
+            tstart = time.time()
+            self.conn.wait()
+            self.wait_time += time.time() - tstart
+            self.current_ids = cast(Optional[DictCursorResults],
+                                    self.conn.cursor.fetchall())
 
-    def __init__(self, rank):
-        self.rank = rank
+        return self.current_ids if self.current_ids is not None else []
 
-    def name(self):
-        return "boundaries rank {}".format(self.rank)
+    def __enter__(self) -> 'PlaceFetcher':
+        return self
 
-    def sql_count_objects(self):
-        return """SELECT count(*) FROM placex
-                  WHERE indexed_status > 0
-                    AND rank_search = {}
-                    AND class = 'boundary' and type = 'administrative'""".format(self.rank)
 
-    def sql_get_objects(self):
-        return """SELECT place_id FROM placex
-                  WHERE indexed_status > 0 and rank_search = {}
-                        and class = 'boundary' and type = 'administrative'
-                  ORDER BY partition, admin_level""".format(self.rank)
+    def __exit__(self, exc_type: Any, exc_value: Any, traceback: Any) -> None:
+        assert self.conn is not None
+        self.conn.wait()
+        self.close()
 
-    @staticmethod
-    def sql_index_place(ids):
-        return "UPDATE placex SET indexed_status = 0 WHERE place_id IN ({})"\
-               .format(','.join((str(i) for i in ids)))
 
 class Indexer:
     """ Main indexing routine.
     """
 
-    def __init__(self, dsn, num_threads):
-        self.conn = psycopg2.connect(dsn)
-        self.threads = [DBConnection(dsn) for _ in range(num_threads)]
+    def __init__(self, dsn: str, tokenizer: AbstractTokenizer, num_threads: int):
+        self.dsn = dsn
+        self.tokenizer = tokenizer
+        self.num_threads = num_threads
+
+
+    def has_pending(self) -> bool:
+        """ Check if any data still needs indexing.
+            This function must only be used after the import has finished.
+            Otherwise it will be very expensive.
+        """
+        with connect(self.dsn) as conn:
+            with conn.cursor() as cur:
+                cur.execute("SELECT 'a' FROM placex WHERE indexed_status > 0 LIMIT 1")
+                return cur.rowcount > 0
+
+
+    def index_full(self, analyse: bool = True) -> None:
+        """ Index the complete database. This will first index boundaries
+            followed by all other objects. When `analyse` is True, then the
+            database will be analysed at the appropriate places to
+            ensure that database statistics are updated.
+        """
+        with connect(self.dsn) as conn:
+            conn.autocommit = True
+
+            def _analyze() -> None:
+                if analyse:
+                    with conn.cursor() as cur:
+                        cur.execute('ANALYZE')
+
+            if self.index_by_rank(0, 4) > 0:
+                _analyze()
+
+            if self.index_boundaries(0, 30) > 100:
+                _analyze()
+
+            if self.index_by_rank(5, 25) > 100:
+                _analyze()
 
-    def index_boundaries(self, minrank, maxrank):
+            if self.index_by_rank(26, 30) > 1000:
+                _analyze()
+
+            if self.index_postcodes() > 100:
+                _analyze()
+
+
+    def index_boundaries(self, minrank: int, maxrank: int) -> int:
+        """ Index only administrative boundaries within the given rank range.
+        """
+        total = 0
         LOG.warning("Starting indexing boundaries using %s threads",
-                    len(self.threads))
+                    self.num_threads)
 
-        for rank in range(max(minrank, 5), min(maxrank, 26)):
-            self.index(BoundaryRunner(rank))
+        with self.tokenizer.name_analyzer() as analyzer:
+            for rank in range(max(minrank, 4), min(maxrank, 26)):
+                total += self._index(runners.BoundaryRunner(rank, analyzer))
 
-    def index_by_rank(self, minrank, maxrank):
-        """ Run classic indexing by rank.
+        return total
+
+    def index_by_rank(self, minrank: int, maxrank: int) -> int:
+        """ Index all entries of placex in the given rank range (inclusive)
+            in order of their address rank.
+
+            When rank 30 is requested then also interpolations and
+            places with address rank 0 will be indexed.
         """
+        total = 0
         maxrank = min(maxrank, 30)
         LOG.warning("Starting indexing rank (%i to %i) using %i threads",
-                    minrank, maxrank, len(self.threads))
+                    minrank, maxrank, self.num_threads)
+
+        with self.tokenizer.name_analyzer() as analyzer:
+            for rank in range(max(1, minrank), maxrank + 1):
+                total += self._index(runners.RankRunner(rank, analyzer), 20 if rank == 30 else 1)
+
+            if maxrank == 30:
+                total += self._index(runners.RankRunner(0, analyzer))
+                total += self._index(runners.InterpolationRunner(analyzer), 20)
 
-        for rank in range(max(1, minrank), maxrank):
-            self.index(RankRunner(rank))
+        return total
 
-        if maxrank == 30:
-            self.index(RankRunner(0))
-            self.index(InterpolationRunner(), 20)
-            self.index(RankRunner(30), 20)
-        else:
-            self.index(RankRunner(maxrank))
 
-    def update_status_table(self):
+    def index_postcodes(self) -> int:
+        """Index the entries of the location_postcode table.
+        """
+        LOG.warning("Starting indexing postcodes using %s threads", self.num_threads)
+
+        return self._index(runners.PostcodeRunner(), 20)
+
+
+    def update_status_table(self) -> None:
         """ Update the status in the status table to 'indexed'.
         """
-        with self.conn.cursor() as cur:
-            cur.execute('UPDATE import_status SET indexed = true')
-        self.conn.commit()
+        with connect(self.dsn) as conn:
+            with conn.cursor() as cur:
+                cur.execute('UPDATE import_status SET indexed = true')
 
-    def index(self, obj, batch=1):
-        """ Index a single rank or table. `obj` describes the SQL to use
+            conn.commit()
+
+    def _index(self, runner: runners.Runner, batch: int = 1) -> int:
+        """ Index a single rank or table. `runner` describes the SQL to use
             for indexing. `batch` describes the number of objects that
             should be processed with a single SQL statement
         """
-        LOG.warning("Starting %s (using batch size %s)", obj.name(), batch)
-
-        cur = self.conn.cursor()
-        cur.execute(obj.sql_count_objects())
+        LOG.warning("Starting %s (using batch size %s)", runner.name(), batch)
 
-        total_tuples = cur.fetchone()[0]
-        LOG.debug("Total number of rows: %i", total_tuples)
+        with connect(self.dsn) as conn:
+            psycopg2.extras.register_hstore(conn)
+            with conn.cursor() as cur:
+                total_tuples = cur.scalar(runner.sql_count_objects())
+                LOG.debug("Total number of rows: %i", total_tuples)
 
-        cur.close()
+            conn.commit()
 
-        progress = ProgressLogger(obj.name(), total_tuples)
+            progress = ProgressLogger(runner.name(), total_tuples)
 
-        if total_tuples > 0:
-            cur = self.conn.cursor(name='places')
-            cur.execute(obj.sql_get_objects())
+            if total_tuples > 0:
+                with conn.cursor(name='places') as cur:
+                    cur.execute(runner.sql_get_objects())
 
-            next_thread = self.find_free_thread()
-            while True:
-                places = [p[0] for p in cur.fetchmany(batch)]
-                if not places:
-                    break
+                    with PlaceFetcher(self.dsn, conn) as fetcher:
+                        with WorkerPool(self.dsn, self.num_threads) as pool:
+                            has_more = fetcher.fetch_next_batch(cur, runner)
+                            while has_more:
+                                places = fetcher.get_batch()
 
-                LOG.debug("Processing places: %s", str(places))
-                thread = next(next_thread)
+                                # asynchronously get the next batch
+                                has_more = fetcher.fetch_next_batch(cur, runner)
 
-                thread.perform(obj.sql_index_place(places))
-                progress.add(len(places))
+                                # And insert the current batch
+                                for idx in range(0, len(places), batch):
+                                    part = places[idx:idx + batch]
+                                    LOG.debug("Processing places: %s", str(part))
+                                    runner.index_places(pool.next_free_worker(), part)
+                                    progress.add(len(part))
 
-            cur.close()
+                            LOG.info("Wait time: fetcher: %.2fs,  pool: %.2fs",
+                                     fetcher.wait_time, pool.wait_time)
 
-            for thread in self.threads:
-                thread.wait()
+                conn.commit()
 
-        progress.done()
-
-    def find_free_thread(self):
-        """ Generator that returns the next connection that is free for
-            sending a query.
-        """
-        ready = self.threads
-        command_stat = 0
-
-        while True:
-            for thread in ready:
-                if thread.is_done():
-                    command_stat += 1
-                    yield thread
-
-            # refresh the connections occasionaly to avoid potential
-            # memory leaks in Postgresql.
-            if command_stat > 100000:
-                for thread in self.threads:
-                    while not thread.is_done():
-                        thread.wait()
-                    thread.connect()
-                command_stat = 0
-                ready = self.threads
-            else:
-                ready, _, _ = select.select(self.threads, [], [])
-
-        assert False, "Unreachable code"
+        return progress.done()