Release Notes




		Version 6.10 of DES adds to previous version (6.9): Enhancements: SQL semantic checking of inconsistent and tautological conditions is extended to dependent views Improved detection of semantic warnings for implied conditions (Error 8 [BG06]) Enhanced SQL semantic detection of unneeded DISTINCT modifier based on functional dependencies and distinguishing if duplicates are enabled or not More SQL hints involving implied conditions Foreign keys can omit the referenced column names, assuming the primary key columns of the referenced table to be the external key Extended SQL CASE for dealing with expressions as arguments Compile-time type checking for arguments of SQL built-ins COALESCE, GREATEST, LEAST, NVL, NVL2, IIF and CASE Supported OFFSET and LIMIT clauses in SQL hints Absorption rules added for conjunction and disjunction, enabling more optimization opportunities Added BIGINT SQL type Open types in messages are now closed Enhanced visibility rules for aliases in SELECT renamings, which allows for references to such aliases in other subsequent clauses Added syntactic error for multiple primary key SQL declarations in the same table Added string to datetime datatype conversion for date without time Support for VALUES in local and global SQL views GROUP BY queries with non-grouped attributes in the SELECT list but determined by groping attributes are accepted With nulls enabled, set operations INTERSECT and EXCEPT (both DISTINCT and ALL) behave as SQL standard dictates. With nulls disabled, more efficient compilations are delivered Differential semi-naïve optimization for linear recursion (/optimize_sn on) is automatically disabled in non-linear rule bodies New commands: /cls Clear the console screen by emitting ANSI escape sequences. This requires a console with ANSI support enabled /sql_semantic_check_data Display whether SQL semantic check based on actual data is enabled (disabled by default). If enabled, data from tables are used for consistency checks and query simplifications /sql_semantic_check_data Switch Enable or disable actual data for SQL semantic check (on or off, resp.) If enabled, data from tables are used for consistency checks and query simplifications Changes: SQL queries that include FROM dual or are FROM-less are now displayed without a FROM clause on DBMS's that support it; in particular, for DES SQL queries that include a top-N clause are now displayed with its simplified version SELECT TOP N on DBMS's that support it; in particular, for DES String types for constants in error messages are not limited to the length of the constant Under strong type casting, the type of a Datalog predicate cannot be changed if its data does not exactly match with the new types The comparison built-ins = and \= respectively use Prolog's =:= and =\= to evaluate its arithmetic arguments, providing a more homogeneous behavior for different Prolog systems. In particular, a Datalog query such as 1=1.0 succeeds for both SICStus and SWI-Prolog host systems Fixed bugs: Some translations of projections to canonical RA in the compilation of SQL queries were omitted. This might lead to compilation errors The operator != was not accepted in DRC Type conversion failed in some corner cases involving disjunctions ('$eq' was not treated as an infix comparison) and string types DWIM could raise an exception in presence of restricted predicates TOP-N semantics was inadvertently changed in version 6.8 for non-open goals. This affected in particular to compilations of SQL EXISTS queries SQL hints should not be issued for TAPI commands except for /parse and /mparse, similar to what semantic errors behave Missing documentation for the modifier ALL of INTERSECT and EXCEPT multiset SQL operations A semantic SQL translation might not terminate because of switched arguments Upon exceptions on an external database, integrity constraint checking might be disabled Numeric types in output schema of queries now identify correct types for constants When detecting tautological conditions and translating to constraints during SQL semantic checking, type casting was enabled in type inference, but it should not have been. Inconsistency checking with type casting enabled may lead to miss integers with floats Opening an external connection (such as SQL Server) may lead to a couple of rows in the dual table, instead just one SQL semantic warning GROUP BY with singleton groups submitted a false positive in two scenarios: for primary keys of more than one column, and cross auto joins. This also led to false positive for replacing GROUP BY by DISTINCT Some SQL queries with ALL/ANY modifiers over subqueries with aggregates raised an exception Aggregates computed in the SELECT list may miss solutions In RA, a rename applied to a sort lead to an unknown column System error exception fixed for SQL query and correlated subquery with ALL and grouping in both Fixed overloaded datetime_sub with type casting on Fixed exception for recompilations of views when switching autocast mode Fixed internal error in derived table with *COUNT() correlated through an expression of an outer attribute Uncorrelated variable in SQL subquery with DISTINCT Replacing a script variable with the same name as a static predicate led to an exception A nested rename in RA might caused input processing error Grouping involving the dual table may return incorrect results An expression involving a scalar SELECT subquery with aggregates led to syntactic error (aggregate in a WHERE) If ORDER BY contained a non-grouped attribute shared with the SELECT list, the error pointed to the first clause instead to the second one Circular aliases may cycle Some relation aliases was missed in subqueries Offset collapsed duplicated entries and thus, in general, less rows than expected were returned From-less SELECT ... INTO statements raised a syntactic error Arithmetic arguments in Datalog string functions were not evaluated A natural full join with no common attributes raised an error Double escape char at the end of the string was misinterpreted In an SQL condition, AND combined with NOT IN raised an error An alias reused in the SELECT list as a single argument raised an error Some remarks inside SQL statements raised a syntax error EXCEPT ALL queries with disabled duplicates did not terminate because different nulls were not identified as a single value in its translation to Datalog's group_by An EXCEPT query involving nulls did not distinguish them as equal values Some constant arguments were passed as positional arguments to GROUP BY The command /drop_tables did not terminate when there were dependency cycles between tables through referential integrity constraints Some types were not closed rightly in view definitions, dealing to incorrect announced types Comparison between string types of different lengths may lead to error even when type casting was enabled Renamings in RA queries might pollute further queries Arguments in ORDER BY (SQL and RA) might be missed with positional An aggregate over a recursive predicate under semi-naïve differential optimization might lead to non-termination The function/predicate IIF assumed true the condition for comparisons with constants Previous versions**