sql.py: mk_insert_select(): INSERT IGNORE: Fixed bug where user-defined types were not supported correctly, by referencing the column's type directly in PL/pgSQL instead of retrieving it from the information_schema. Use a row variable of the output table's row type to store the select query data, so that you don't need to create a separate local var for each output column.

import.stats.xls: Changed \/row (ms)" to "ms/row"

sql.py: mk_insert_select(): INSERT IGNORE: Used typed vars to store each column of the select query because the select query's column names are not known, so it would be unknown which record var column to use for each output column. (Add types so there are no unknown-type values that PostgreSQL will say can't be cast.)

sql.py: insert_select(): In ignore mode, always recover from errors

sql.py: mk_insert_select(): INSERT IGNORE: Implemented simple cursor loop that just inserts each row, without ignoring duplicate keys

sql.py: mk_insert_select(): Added ignore param to do an INSERT IGNORE equivalent operation. Added basic structure for INSERT IGNORE. Support embeddable queries without RETURNING columns (which is the case for most INSERT IGNORE queries).

sql.py: run_query_into(): Always set into.schema to None, so that table is created in the temp schema in debug_temp mode

sql.py: DbConn.DbCursor.execute(): Don't cache ADD COLUMN queries because the rest of the query must be unique in order to use caching

sql.py: add_col(): Don't ignore already-existing columns because sometimes name truncation causes collisions, requiring the caller to version the name. add_index_col(): Version the column name to avoid collisions.

sql.py: add_index(): Version index names to avoid collisions

sql.py: DbConn.DbCursor.execute(): Always cache structural changes, not just if they throw an exception

sql.py: add_index(): Fixed bug where expr could not be deep-copied until ensure_not_null() had been run on it, because ensure_not_null() modifies the index_cols of the table and this needs to be modified on the original table

sql.py: add_index(): Fixed bug where expr could not be deep-copied until ensure_not_null() had been run on it, because ensure_not_null() modifies the index_cols of the table and this needs to be modified on the original table

sql_gen.py: EnsureNotNull: Added to_str() which uses the column's index col if available. This works around a PostgreSQL bug where indexes on expressions are sometimes not used, but indexes on plain columns are used.

sql_gen.py: Store index cols in the table instead of in the column, because columns are often recreated from the table and name but tables are generally not copied

sql_gen.py: Added is_indexed_col()

sql.py: add_index(): Use new sql.ensure_not_null(), which creates a separate column to store the index expr where possible

sql.py: Added add_not_null(), add_index_col(), ensure_not_null()

sql_gen.py: Added is_temp_col()

sql_gen.py: concat(): Also preserve cast suffix

sql_gen.py: Col: Added index_col attr

sql_gen.py: Added suffixed_col()

sql.py: run_query_into(): Set the into table to be a temp table

sql_gen.py: Table: Store whether table is temp table

sql_gen.py: concat(): Also preserve appended column names

sql_gen.py: Col.to_str(): Take for_str param which does both concat() and str()/clean_name() instead of use_concat param which doesn't remove quotes from the strings before concatenating, causing strings to be incorrectly truncated

sql_gen.py: EnsureNotNull: Cast the null sentinel to the type it's the sentinel for, so that COALESCE returns the same type as the provided type, instead of potentially incorrectly assuming the constant's type. This fixes a bug where COALESCE returned an incorrect type.

sql.py: cast(): Use new sql_gen.Cast

sql_gen.py: Added Cast

sql_gen.py: concat(): Consider ")"s part of the existing suffix, so that parentheses match up

Removed extra trailing whitespace

main Makefile: Added explicit schemas/temp/reinstall target because vegbiendev (Ubuntu 12.04) seems to ignore %/reinstall

vegbien_dest: Added temp to schemas list so the intermediate tables will be created there, where they are easy to remove with `make schemas/temp/reinstall`

main Makefile: Added %/reinstall, which calls the corresponding uninstall and install targets

Added temp schema to store intermediate tables permanently when desired for debugging. It is not part of schemas/all because it should not be remade to include current intermediate tables when remaking other schema exports.

schemas: Moved *.sql.make into Makefile to take advantage of % pattern matching

Removed no longer used schemas/vegbien_empty.sql

main Makefile: Removed empty_db, because `make schemas/reinstall` has the same effect and is simpler

README.TXT: Changed documentation to use make schemas/reinstall to empty the DB, since that command is simpler. Added how to archive the last import.

db_xml.py: put_table(): Removed `if not db.debug_temp` check because that is done by sql.empty_temp()

sql.py: put_table(): Use new empty_temp()

import.stats.xls: Added comments for estimated numbers. Added "," separators to large numbers.

sql.py: empty_temp(): In debug_temp mode, leave temp tables there for debugging

sql.py: empty_temp(): Don't output at log_level 2 because it's an internal query, not part of the core algorithm

sql.py: truncate(): Added kw_args to pass to run_query()

Added inputs/import.stats.xls, which compares row-based and column-based import. This shows that column-based import is slowed down by table locking when run simultaneously, so we will need a new INSERT IGNORE replacement that doesn't lock tables.

inputs: Ignore OpenOffice.org lock files

sql.py: empty_temp(): Don't print log message if not emptying any tables

db_xml.py: put_table(): Empty unneeded temp tables to free up memory

sql.py: Added empty_temp()

sql.py: Use new lists.mk_seq()

lists.py: Added mk_seq()

lists.py: is_seq(): Also return true for sets

schemas/vegbien.sql: specimenreplicate: Added indexes using COALESCE to match what sql_gen does

sql.py: put_table(): Getting output table pkeys of existing/inserted rows: Do a DISTINCT ON the input pkey (row_num) in case the plain JOIN matched multiple output table rows for one input table row

sql.py: put_table(): Empty unneeded temp tables to free up memory and avoid running out of memory (the temp tables seem to be in-memory only)

sql_gen.py: null_sentinels: Added value for type timestamp with time zone. Put each type on its own line for clarity.

schemas/vegbien.sql: locationdetermination: Changed indexes to use COALESCE to match what sql_gen now does

schemas/vegbien.sql: location: Added indexes using COALESCE to match what sql_gen does

sql.py: cast_temp_col(): Add an index on the created column

sql_gen.py: null_sentinels: Added value for type double precision

sql_gen.py: ensure_not_null(): Warn of no null sentinel for type, even if caller catches error

schemas/py_functions.sql: Added plain function _namePart() and use it in trigger function _namePart()

schemas/py_functions.sql: Added plain functions _dateRangeStart() and _dateRangeEnd() and use them in trigger functions _dateRangeStart() and _dateRangeEnd()

schemas/functions.sql: _label(): Ensure that label is NOT NULL so it doesn't NULL out the entire string

schemas/functions.sql: Added plain function _nullIf() and use it in trigger function _nullIf()

sql.py: DbConn.DbCursor._cache_result(): Corrected comment to reflect why different types of queries are cached differently

sql.py: add_col(): Catch DuplicateExceptions so that columns that already existed are ignored

sql.py: run_query(): DuplicateException: Also match "column already exists" errors

sql.py: Merged DuplicateTableException and DuplicateFunctionException into one exception DuplicateException, with a type variable for the type of duplicate item. Added ExceptionWithNameType.

schemas/functions.sql: Fixed bug where external function calls needed to be schema-qualified in case functions schema is not in the search_path

schemas/functions.sql: Added plain function _label() and use it in trigger function _label()

sql.py: put_table(): Support plain SQL functions in addition to relational functions

sql_gen.py: Added NamedArg. FunctionCall: Support named arguments (http://www.postgresql.org/docs/9.0/static/sql-syntax-calling-funcs.html).

schemas/functions.sql: Added plain function _merge() and use it in trigger function _merge()

schemas/functions.sql: Added plain function _alt() and use it in trigger function _alt()

schemas/functions.sql: Removed no longer used ensure_not_null()

sql.py: put_table(): MissingCastException: Use cast_temp_col() so that cast will occur before any main insert, which locks the output table and should take as little time as possible

sql.py: Added cast_temp_col()

sql.py: add_col(): Support additional run_query() kw_args. add_row_num(): Use new add_col().

sql.py: Added add_col()

sql_gen.py: Col.__str__(): Truncate any table name using concat() to ensure that the full column name is included in the string

strings.py, sql_gen.py: Renamed add_suffix() to concat() to reflect that this is a fixed-length replacement for +

sql.py: put_table(): Moved MissingCastException to the top of the exceptions list because it's more of a core exception than the others, and will be raised before any rows are even inserted

sql.py: DbConn.with_savepoint(): Always release savepoint, because after ROLLBACK TO SAVEPOINT, "The savepoint remains valid and can be rolled back to again" (http://www.postgresql.org/docs/8.3/static/sql-rollback-to.html). Moved `self._savepoint -= 1` to the main try block's new finally block.

sql.py: put_table(): Lock output table right before, and in the same nested transaction as, the insert statement that needs lock, so that it is not released in a prior autocommit and is held for as little time as possible

db_xml.py: put_table(): Removed no longer needed commit param

bin/map: Removed rollback() call before closing the connection because PostgreSQL does this automatically

csv2db: Log inserts with log_level=5 so they are not shown for verbosity 4, which is used to see the savepoints and autocommits

Removed unnecessary db.db.commit() calls because commits are now done automatically by DbConn's autocommit mode

sql.py: DbConn.do_autocommit(): Output the "Autocommitting" debug message with level=4 so that it doesn't clutter up the logging output for normal verbosities

DbConn: autocommit mode defaults to True so that all scripts get the benefit of automatic commits

input.Makefile: Staging tables: import/install-%: Include the table name in the log file name so that successive tables for the same datasource don't overwrite the same log file

sql.py: DbConn: Don't always autocommit in debug_temp mode, because this could cause autocommit mode to be turned on when the user does not expect it

bin/map: connect_db(): Autocommit in commit mode to avoid the need for manual commits. This should also reduce the time that table locks are held, to avoid unnecessary contention when multiple processes are trying to insert into the same output table. (The program always uses nested transactions to support rollbacks, so there is no problem autocommitting whenever a top-level nested transaction or top-level query completes.)

sql_gen.py: Removed TempFunction because that functionality is now provided by DbConn.TempFunction()

sql.py: Use new DbConn.TempFunction()

sql.py: DbConn: Added TempFunction()

sql.py: Use new DbConn.debug_temp config option to control whether temporary objects should instead be permanent

sql.py: DbConn: Added config option debug_temp