load: Add --recursive option.

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

diff --git a/pokedex/db/load.py b/pokedex/db/load.py
index 4512bf6..88f5332 100644 (file)
--- a/pokedex/db/load.py
+++ b/pokedex/db/load.py
@@ -11,6 +11,7 @@ import sqlalchemy.types
 from pokedex.db import metadata
 import pokedex.db.tables as tables
 from pokedex.defaults import get_default_csv_dir
+from pokedex.db.dependencies import find_dependent_tables
 
 
 def _get_table_names(metadata, patterns):
@@ -95,7 +96,7 @@ def _get_verbose_prints(verbose):
     return print_start, print_status, print_done
 
 
-def load(session, tables=[], directory=None, drop_tables=False, verbose=False, safe=True):
+def load(session, tables=[], directory=None, drop_tables=False, verbose=False, safe=True, recursive=False):
     """Load data from CSV files into the given database session.
 
     Tables are created automatically.
@@ -119,6 +120,9 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False, s
     `safe`
         If set to False, load can be faster, but can corrupt the database if
         it crashes or is interrupted.
+
+    `recursive`
+        If set to True, load all dependent tables too.
     """
 
     # First take care of verbosity
@@ -128,8 +132,13 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False, s
     if directory is None:
         directory = get_default_csv_dir()
 
+    # XXX why isn't this done in command_load
     table_names = _get_table_names(metadata, tables)
     table_objs = [metadata.tables[name] for name in table_names]
+
+    if recursive:
+        table_objs.extend(find_dependent_tables(table_objs))
+
     table_objs = sqlalchemy.sql.util.sort_tables(table_objs)
 
     # SQLite speed tweaks
@@ -140,12 +149,16 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False, s
     # Drop all tables if requested
     if drop_tables:
         print_start('Dropping tables')
-        for table in reversed(table_objs):
+        for n, table in enumerate(reversed(table_objs)):
             table.drop(checkfirst=True)
+            print_status('%s/%s' % (n, len(table_objs)))
         print_done()
 
-    for table in table_objs:
+    print_start('Creating tables')
+    for n, table in enumerate(table_objs):
         table.create()
+        print_status('%s/%s' % (n, len(table_objs)))
+    print_done()
     connection = session.connection()
 
     # Okay, run through the tables and actually load the data now
@@ -170,14 +183,10 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False, s
 
         if not safe and session.connection().dialect.name == 'postgresql':
             """
-            Postgres' CSV dialect is nearly the same as ours, except that it
-            treats completely empty values as NULL, and empty quoted
-            strings ("") as an empty strings.
-            Pokedex dump does not quote empty strings. So, both empty strings
-            and NULLs are read in as NULL.
-            For an empty string in a NOT NULL column, the load will fail, and
-            load will fall back to the cross-backend row-by-row loading. And in
-            nullable columns, we already load empty stings as NULL.
+            Postgres' CSV dialect works with our data, if we mark the not-null
+            columns with FORCE NOT NULL.
+            COPY is only allowed for DB superusers. If you're not one, use safe
+            loading (pokedex load -S).
             """
             session.commit()
             not_null_cols = [c for c in column_names if not table_obj.c[c].nullable]
@@ -185,15 +194,15 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False, s
                 force_not_null = 'FORCE NOT NULL ' + ','.join('"%s"' % c for c in not_null_cols)
             else:
                 force_not_null = ''
-            command = "COPY {table_name} ({columns}) FROM '{csvpath}' CSV HEADER {force_not_null}"
+            command = "COPY %(table_name)s (%(columns)s) FROM '%(csvpath)s' CSV HEADER %(force_not_null)s"
             session.connection().execute(
-                    command.format(
-                            table_name=table_name,
-                            csvpath=csvpath,
-                            columns=','.join('"%s"' % c for c in column_names),
-                            force_not_null=force_not_null,
-                        )
+                command % dict(
+                    table_name=table_name,
+                    csvpath=csvpath,
+                    columns=','.join('"%s"' % c for c in column_names),
+                    force_not_null=force_not_null,
                 )
+            )
             session.commit()
             print_done()
             continue
@@ -213,6 +222,8 @@ def load(session, tables=[], directory=None, drop_tables=False, verbose=False, s
 
         new_rows = []
         def insert_and_commit():
+            if not new_rows:
+                return
             session.connection().execute(insert_stmt, new_rows)
             session.commit()
             new_rows[:] = []