]> git.openstreetmap.org Git - nominatim.git/blobdiff - nominatim/indexer/indexer.py
more formatting fixes
[nominatim.git] / nominatim / indexer / indexer.py
index b7673abaddc8090a896351c7ad230f372742a739..d0cfb391c4dbdf7a63c875af6ec1b2d98ca88d0c 100644 (file)
@@ -2,14 +2,13 @@
 Main work horse for indexing (computing addresses) the database.
 """
 import logging
-import select
 import time
 
 import psycopg2.extras
 
 from nominatim.indexer.progress import ProgressLogger
 from nominatim.indexer import runners
-from nominatim.db.async_connection import DBConnection
+from nominatim.db.async_connection import DBConnection, WorkerPool
 from nominatim.db.connection import connect
 
 LOG = logging.getLogger()
@@ -81,73 +80,6 @@ class PlaceFetcher:
         self.conn.wait()
         self.close()
 
-class WorkerPool:
-    """ A pool of asynchronous database connections.
-
-        The pool may be used as a context manager.
-    """
-    REOPEN_CONNECTIONS_AFTER = 100000
-
-    def __init__(self, dsn, pool_size):
-        self.threads = [DBConnection(dsn) for _ in range(pool_size)]
-        self.free_workers = self._yield_free_worker()
-        self.wait_time = 0
-
-
-    def finish_all(self):
-        """ Wait for all connection to finish.
-        """
-        for thread in self.threads:
-            while not thread.is_done():
-                thread.wait()
-
-        self.free_workers = self._yield_free_worker()
-
-    def close(self):
-        """ Close all connections and clear the pool.
-        """
-        for thread in self.threads:
-            thread.close()
-        self.threads = []
-        self.free_workers = None
-
-
-    def next_free_worker(self):
-        """ Get the next free connection.
-        """
-        return next(self.free_workers)
-
-
-    def _yield_free_worker(self):
-        ready = self.threads
-        command_stat = 0
-        while True:
-            for thread in ready:
-                if thread.is_done():
-                    command_stat += 1
-                    yield thread
-
-            if command_stat > self.REOPEN_CONNECTIONS_AFTER:
-                for thread in self.threads:
-                    while not thread.is_done():
-                        thread.wait()
-                    thread.connect()
-                ready = self.threads
-                command_stat = 0
-            else:
-                tstart = time.time()
-                _, ready, _ = select.select([], self.threads, [])
-                self.wait_time += time.time() - tstart
-
-
-    def __enter__(self):
-        return self
-
-
-    def __exit__(self, exc_type, exc_value, traceback):
-        self.finish_all()
-        self.close()
-
 
 class Indexer:
     """ Main indexing routine.
@@ -160,7 +92,7 @@ class Indexer:
 
 
     def index_full(self, analyse=True):
-        """ Index the complete database. This will first index boudnaries
+        """ 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.
@@ -168,13 +100,10 @@ class Indexer:
         with connect(self.dsn) as conn:
             conn.autocommit = True
 
-            if analyse:
-                def _analyze():
+            def _analyze():
+                if analyse:
                     with conn.cursor() as cur:
                         cur.execute('ANALYZE')
-            else:
-                def _analyze():
-                    pass
 
             self.index_by_rank(0, 4)
             _analyze()
@@ -274,7 +203,7 @@ class Indexer:
 
                                 # And insert the curent batch
                                 for idx in range(0, len(places), batch):
-                                    part = places[idx:idx+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))