sql_io.py: import_csv(): Only do the import in a savepoint if using COPY FROM, to allow autocommits after each insert and thus make rows visible immediately after they are inserted

db_xml.py: put_table(): Subsetting in_table: Add a row number column if in_table does not already have a pkey

db_xml.py: put_table(): Subsetting in_table: Copy all of in_table's structure, rather than just the column types, by using sql.copy_table_struct() and sql.insert_select(). This preserves pkeys and NOT NULL constraints, which are useful for column-based import.

db_xml.py: put_table(): Subsetting in_table: Create in_table as a completely new sql_gen.Table instead of copying full_in_table and relying on sql.run_query_into() to set is_temp and remove the schema

sql.py: add_row_num(): Use if_not_exists in order to abort if the column already exists rather than adding a version #

sql.py: add_col(): Added if_not_exists param to abort if the column already exists rather than adding a version #

db_xml.py: put_table(): Removed no longer accurate comment that full_in_table will be shadowed (hidden) by the created temp table. (The temp table is now named differently, so the shadowing does not occur.)

db_xml.py: put_table(): Replaced no longer accurate Recurse comment with Import data. Rewrapped lines.

sql_io.py: import_csv(): Factored insertion code out into new append_csv()

README.TXT: Data import: `make test by_col=1`: Replaced errors explanation with pointer to updated explanation in the Testing section