Speed tweaks for pokedex load in SQLite

[zzz-pokedex.git] / pokedex / db / load.py

diff --git a/pokedex/db/load.py b/pokedex/db/load.py
index e3e7fa5..e6b3287 100644 (file)
--- a/pokedex/db/load.py
+++ b/pokedex/db/load.py
@@ -1,8 +1,7 @@
 """CSV to database or vice versa."""
 import csv
+import fnmatch
 import os.path
-import pkg_resources
-import re
 import sys
 
 from sqlalchemy.orm.attributes import instrumentation_registry
@@ -11,38 +10,27 @@ import sqlalchemy.types
 
 from pokedex.db import metadata
 import pokedex.db.tables as tables
+from pokedex.defaults import get_default_csv_dir
 
 
-def _wildcard_char_to_regex(char):
-    """Converts a single wildcard character to the regex equivalent."""
-
-    if char == '?':
-        return '.?'
-    elif char == '*':
-        return '.*'
+def _get_table_names(metadata, patterns):
+    """Returns a list of table names from the given metadata.  If `patterns`
+    exists, only tables matching one of the patterns will be returned.
+    """
+    if patterns:
+        table_names = set()
+        for pattern in patterns:
+            if '.' in pattern or '/' in pattern:
+                # If it looks like a filename, pull out just the table name
+                _, filename = os.path.split(pattern)
+                table_name, _ = os.path.splitext(filename)
+                pattern = table_name
+
+            table_names.update(fnmatch.filter(metadata.tables.keys(), pattern))
     else:
-        return re.escape(char)
-
-def _wildcard_glob_to_regex(glob):
-    """Converts a single wildcard glob to a regex STRING."""
-
-    # If it looks like a filename, make it not one
-    if '.' in glob or '/' in glob:
-        _, filename = os.path.split(glob)
-        table_name, _ = os.path.splitext(filename)
-        glob = table_name
-
-    return u''.join(map(_wildcard_char_to_regex, glob))
-
-def _wildcards_to_regex(strings):
-    """Converts a list of wildcard globs to a single regex object."""
-
-    regex_parts = map(_wildcard_glob_to_regex, strings)
-
-    regex = '^(?:' + '|'.join(regex_parts) + ')$'
-
-    return re.compile(regex)
+        table_names = metadata.tables.keys()
 
+    return list(table_names)
 
 def _get_verbose_prints(verbose):
     """If `verbose` is true, returns three functions: one for printing a
@@ -107,7 +95,7 @@ def _get_verbose_prints(verbose):
     return print_start, print_status, print_done
 
 
-def load(session, tables=[], directory=None, drop_tables=False, verbose=False):
+def load(session, tables=[], directory=None, drop_tables=False, verbose=False, safe=True):
     """Load data from CSV files into the given database session.
 
     Tables are created automatically.
@@ -127,24 +115,27 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False):
 
     `verbose`
         If set to True, status messages will be printed to stdout.
+
+    `safe`
+        If set to False, load can be faster, but can corrupt the database if
+        it crashes or is interrupted.
     """
 
     # First take care of verbosity
     print_start, print_status, print_done = _get_verbose_prints(verbose)
 
 
-    if not directory:
-        directory = pkg_resources.resource_filename('pokedex', 'data/csv')
-
-    if tables:
-        regex = _wildcards_to_regex(tables)
-        table_names = filter(regex.match, metadata.tables.keys())
-    else:
-        table_names = metadata.tables.keys()
+    if directory is None:
+        directory = get_default_csv_dir()
 
+    table_names = _get_table_names(metadata, tables)
     table_objs = [metadata.tables[name] for name in table_names]
     table_objs = sqlalchemy.sql.util.sort_tables(table_objs)
 
+    # SQLite speed tweaks
+    if not safe and session.connection().dialect.name == 'sqlite':
+        session.connection().execute("PRAGMA synchronous=OFF")
+        session.connection().execute("PRAGMA journal_mode=OFF")
 
     # Drop all tables if requested
     if drop_tables:
@@ -196,7 +187,7 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False):
             session.commit()
             new_rows[:] = []
 
-            progress = "{0}%".format(100 * csvfile.tell() // csvsize)
+            progress = "%d%%" % (100 * csvfile.tell() // csvsize)
             print_status(progress)
 
         for csvs in reader:
@@ -268,6 +259,10 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False):
 
         print_done()
 
+    # SQLite check
+    if session.connection().dialect.name == 'sqlite':
+        session.connection().execute("PRAGMA integrity_check")
+
 
 
 def dump(session, tables=[], directory=None, verbose=False):
@@ -293,14 +288,9 @@ def dump(session, tables=[], directory=None, verbose=False):
 
 
     if not directory:
-        directory = pkg_resources.resource_filename('pokedex', 'data/csv')
-
-    if tables:
-        regex = _wildcards_to_regex(tables)
-        table_names = filter(regex.match, metadata.tables.keys())
-    else:
-        table_names = metadata.tables.keys()
+        directory = get_default_csv_dir()
 
+    table_names = _get_table_names(metadata, tables)
     table_names.sort()