]> git.openstreetmap.org Git - nominatim.git/blobdiff - src/nominatim_db/indexer/indexer.py
disable ZIP5+ test for Python frontend
[nominatim.git] / src / nominatim_db / indexer / indexer.py
index 5a219f6b2ecc3904b1129c9b4f9cff92c12132a5..9680e5a9db14d382e84f385c7ffeaf6cb1c3dd13 100644 (file)
@@ -7,92 +7,20 @@
 """
 Main work horse for indexing (computing addresses) the database.
 """
-from typing import Optional, Any, cast
+from typing import cast, List, Any
 import logging
 import time
 
-import psycopg2.extras
+import psycopg
 
-from ..typing import DictCursorResults
-from ..db.async_connection import DBConnection, WorkerPool
-from ..db.connection import connect, Connection, Cursor
+from ..db.connection import connect, execute_scalar
+from ..db.query_pool import QueryPool
 from ..tokenizer.base import AbstractTokenizer
 from .progress import ProgressLogger
 from . import runners
 
 LOG = logging.getLogger()
 
-
-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 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.
-
-            Returns true if there is still data available.
-        """
-        ids = cast(Optional[DictCursorResults], cur.fetchmany(100))
-
-        if not ids:
-            self.current_ids = None
-            return False
-
-        assert self.conn is not None
-        self.current_ids = runner.get_place_details(self.conn, ids)
-
-        return True
-
-    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
-
-        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())
-
-        return self.current_ids if self.current_ids is not None else []
-
-    def __enter__(self) -> 'PlaceFetcher':
-        return self
-
-
-    def __exit__(self, exc_type: Any, exc_value: Any, traceback: Any) -> None:
-        assert self.conn is not None
-        self.conn.wait()
-        self.close()
-
-
 class Indexer:
     """ Main indexing routine.
     """
@@ -114,7 +42,7 @@ class Indexer:
                 return cur.rowcount > 0
 
 
-    def index_full(self, analyse: bool = True) -> None:
+    async 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
@@ -128,23 +56,27 @@ class Indexer:
                     with conn.cursor() as cur:
                         cur.execute('ANALYZE')
 
-            if self.index_by_rank(0, 4) > 0:
-                _analyze()
+            while True:
+                if await self.index_by_rank(0, 4) > 0:
+                    _analyze()
+
+                if await self.index_boundaries(0, 30) > 100:
+                    _analyze()
 
-            if self.index_boundaries(0, 30) > 100:
-                _analyze()
+                if await self.index_by_rank(5, 25) > 100:
+                    _analyze()
 
-            if self.index_by_rank(5, 25) > 100:
-                _analyze()
+                if await self.index_by_rank(26, 30) > 1000:
+                    _analyze()
 
-            if self.index_by_rank(26, 30) > 1000:
-                _analyze()
+                if await self.index_postcodes() > 100:
+                    _analyze()
 
-            if self.index_postcodes() > 100:
-                _analyze()
+                if not self.has_pending():
+                    break
 
 
-    def index_boundaries(self, minrank: int, maxrank: int) -> int:
+    async def index_boundaries(self, minrank: int, maxrank: int) -> int:
         """ Index only administrative boundaries within the given rank range.
         """
         total = 0
@@ -153,11 +85,11 @@ class Indexer:
 
         with self.tokenizer.name_analyzer() as analyzer:
             for rank in range(max(minrank, 4), min(maxrank, 26)):
-                total += self._index(runners.BoundaryRunner(rank, analyzer))
+                total += await self._index(runners.BoundaryRunner(rank, analyzer))
 
         return total
 
-    def index_by_rank(self, minrank: int, maxrank: int) -> int:
+    async 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.
 
@@ -171,21 +103,27 @@ class Indexer:
 
         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 rank >= 30:
+                    batch = 20
+                elif rank >= 26:
+                    batch = 5
+                else:
+                    batch = 1
+                total += await self._index(runners.RankRunner(rank, analyzer), batch)
 
             if maxrank == 30:
-                total += self._index(runners.RankRunner(0, analyzer))
-                total += self._index(runners.InterpolationRunner(analyzer), 20)
+                total += await self._index(runners.RankRunner(0, analyzer))
+                total += await self._index(runners.InterpolationRunner(analyzer), 20)
 
         return total
 
 
-    def index_postcodes(self) -> int:
+    async 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)
+        return await self._index(runners.PostcodeRunner(), 20)
 
 
     def update_status_table(self) -> None:
@@ -197,46 +135,58 @@ class Indexer:
 
             conn.commit()
 
-    def _index(self, runner: runners.Runner, batch: int = 1) -> int:
+    async 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)", runner.name(), batch)
 
-        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)
-
-            conn.commit()
+        total_tuples = self._prepare_indexing(runner)
 
-            progress = ProgressLogger(runner.name(), total_tuples)
+        progress = ProgressLogger(runner.name(), total_tuples)
 
-            if total_tuples > 0:
-                with conn.cursor(name='places') as cur:
-                    cur.execute(runner.sql_get_objects())
+        if total_tuples > 0:
+            async with await psycopg.AsyncConnection.connect(
+                                 self.dsn, row_factory=psycopg.rows.dict_row) as aconn,\
+                       QueryPool(self.dsn, self.num_threads, autocommit=True) as pool:
+                fetcher_time = 0.0
+                tstart = time.time()
+                async with aconn.cursor(name='places') as cur:
+                    query = runner.index_places_query(batch)
+                    params: List[Any] = []
+                    num_places = 0
+                    async for place in cur.stream(runner.sql_get_objects()):
+                        fetcher_time += time.time() - tstart
 
-                    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()
+                        params.extend(runner.index_places_params(place))
+                        num_places += 1
 
-                                # asynchronously get the next batch
-                                has_more = fetcher.fetch_next_batch(cur, runner)
+                        if num_places >= batch:
+                            LOG.debug("Processing places: %s", str(params))
+                            await pool.put_query(query, params)
+                            progress.add(num_places)
+                            params = []
+                            num_places = 0
 
-                                # 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))
+                        tstart = time.time()
 
-                            LOG.info("Wait time: fetcher: %.2fs,  pool: %.2fs",
-                                     fetcher.wait_time, pool.wait_time)
+                if num_places > 0:
+                    await pool.put_query(runner.index_places_query(num_places), params)
 
-                conn.commit()
+            LOG.info("Wait time: fetcher: %.2fs,  pool: %.2fs",
+                     fetcher_time, pool.wait_time)
 
         return progress.done()
+
+
+    def _prepare_indexing(self, runner: runners.Runner) -> int:
+        with connect(self.dsn) as conn:
+            hstore_info = psycopg.types.TypeInfo.fetch(conn, "hstore")
+            if hstore_info is None:
+                raise RuntimeError('Hstore extension is requested but not installed.')
+            psycopg.types.hstore.register_hstore(hstore_info)
+
+            total_tuples = execute_scalar(conn, runner.sql_count_objects())
+            LOG.debug("Total number of rows: %i", total_tuples)
+        return cast(int, total_tuples)