Skip to content

Version 21.1
Compare
Choose a tag to compare
@darold darold released this 01 Apr 20:39
· 779 commits to master since this release
v21.1
8485715

2021 04 01 - v21.1

This release fix several issues reported since past six months and adds some new features and improvements.

  • Now that Orafce 3.15.0 has a definition for the REGEXP_* function, makes the translation optional to USE_ORAFCE directive.
  • Add set application name in connection to Oracle/MySql/PostgreSQL.
  • Add translation of REGEXP_COUNT() and change assessment cost.
  • Rewrite the way REGEXP_LIKE() is translated into regexp_match to support modifiers. This rewrite also fix default behavior between
    Oracle and PostgreSQL.
  • Replace DBMS_LOB.GETLENGTH() by PostgreSQL octet_length() function.
  • Add types correspondences for VARCHAR2 and NVARCHAR2 in DATA_TYPE configuration directive.
  • Add autodetection and support of geometry type, srid and dimension for ArcGis geometries.
  • Add conversion of default value in function parameters.
  • Add -u | --unit option to ora2pg_scanner to be able to set the migration cost unit value globally.
  • Replace DBMS_LOB.SUBSTR() by SUBSTR()
  • Remove TO_CLOB() it is useless, manual cast could be necessary.
  • Replace IS JSON validation clause in CHECK constraints by (CASE WHEN $1::json IS NULL THEN true ELSE true END). When the code is invalid an error is fired.
  • DISTINCT and UNIQUE are synonym on Oracle.

Backward compatibility changes:

  • Force a column to be bigint if this is an identity column. Thanks to MigOps.com for the patch.
  • Fix EMPTY_LOB_NULL, enable/disable was inverted, keeping default to enabled. Take care that in old ora2pg.conf it is disabled so it
    will break backward compatibility with old configuration.
  • Replace NO_LOB_LOCATOR with USE_LOB_LOCATOR and NO_BLOB_EXPORT with ENABLE_BLOB_EXPORT to avoid confusion with double negative variable. Backward compatibility is preserved with a warning.
  • SRID for SDO_GEOMETRY export is now taken from the value not forced from the metadata table.

Here is the full list of changes and acknowledgements:

  - Take Geometry SRID from the data and fallback to SRID defined in
    metadata when not found. Thanks to Sebastian Albert for the report.
  - Fix case where Ora2Pg temporary substitution of '' by placeholder
    was not restored. Thanks to MigOps.com for the patch.
  - Fix identity column export on unsupported Oracle 18c options.
    Thanks to MigOps.com for the patch.
  - Fix export of columns indexes created with single quote.
    Thanks to MigOps.com for the patch.
  - Fix replacement of keyword PROCEDURE by FUNCTION in constraints
    constants definition. Thanks to marie-joechahine for the report.
  - Replace IS JSON validation clause in CHECK constraints. Thanks to
    marie-joechahine for the report and MigOps.com for the patch.
  - Add support to ON OVERFLOW clause in LISTAGG replacement.
    Thanks to MigOps.com for the patch.
  - Fix incorrect handling of HAVING+GROUP BY rewriting.
    Thanks to MigOps.com for the patch.
  - Add replacement of TO_NCHAR by a cast to varchar. Thanks to
    MigOps.com for the patch.
  - Fix replacement of NOTFOUND when there is extra space or new line
    in the WHEN EXIT clause. Thanks to MigOps.com for the patch.
  - Fix a regression in NO_VIEW_ORDERING, it was not taken in account
    anymore. Thanks to RonJojn2 for the report.
  - Replace DATA_TYPE with DTD_IDENTIFIER in MySQL catalog queries for
    version prior 5.5.0. Thanks to zejeanmi for the report.
  - Fix import script to import sequences before tables. Thanks to
    MigOps.com for the patch.
  - Fix detail report of custom type in migration assessment. Thanks
    to MigOps.com for the patch.
  - Fix duplicate schema prefixed to SYNONYM. Thanks to dlc75 for the
    reports.
  - Replace NO_LOB_LOCATOR with USE_LOB_LOCATOR and NO_BLOB_EXPORT with
    ENABLE_BLOB_EXPORT to avoid confusion with double negative variable.
    Thanks to Rob Johnson for the report.
  - Fix some missing replacements of NVL and rewrite !=-1 into != -1.
    Thanks to MigOps.com for the patch.
  - Fix ROWNUM followed by + or - operator and when no aliases are
    provided. Thanks to MigOps.com for the patch.
  - Add DBSFWUSER to the list of user/schema exclusion. Thanks to
    MigOps.com for the patch.
  - Fix regexp to not append subquery aliases on JOIN clause. Thanks
    to Rui Pereira for the report.
  - Handle PRESERVE_CASE and EXPORT_SCHEMA in sequence name. Thanks
    to marie-joechahine for the report.
  - Add CREATE SCHEMA statement to sequence export when EXPORT_SCHEMA
    is enabled. Thanks to marie-joechahine for the report.
  - Fix duplicate index name on subpartition. Thanks to Philippe
    Beaudoin for the report.
  - Exclude sequences used for IDENTITY column (ISEQ$$_). Thanks to
    marie-joechahine for the report.
  - Fix parsing from file of CREATE SEQUENCE. Thanks to Rui Pereira
    for the report.
  - In export_all.sh script use the database owner provided if it is a
    superuser instead of postgres user. Thanks to jjune235 for the
    feature request.
  - Fix parsing of triggers when there is a CASE inside the code.
    Thanks to Rui Pereira for the report.
  - Add set application name in connection to Oracle/MySql/PostgreSQL.
    Thanks to Yoni Sade for the patch.
  - Fix double column alias when replacing ROWNUM. Thanks to Rui
    Pereira for the report.
  - Add translation of the REGEXP_COUNT function and change assessment
    cost.
  - Rewrite the way REGEXP_LIKE is translated into regexp_match to
    support modifiers. This rewrite also fix default behavior between
    Oracle and PostgreSQL.  Thanks to otterrisk for the report.
  - Add IS JSON to assessment. Thanks to marie-joe Chahine for the
    report.
  - Fix multi-columns RANGE partitioning. Thanks to Philippe Beaudoin
    for the report.
  - Improve reordering columns. Sort by fieldsize first, if same size
    then it sorts by original position. Thanks to Sebastien Caunes for
    the patch.
  - Append partition's column to the primary key of the table as it
    must be part of the PK on PostgreSQL. Thanks to xinjirufen for the
    report.
  - Fix partition export where PRESERVE_CASE was applied to Oracle
    side. Thanks to schleb1309 for the report.
  - Fix trigger export with column restriction. Thanks to Sebastien
    Caunes for the report.
  - Update installation information.
  - Fix table reordering following data type. Thanks to Sebastien
    Caunes for the patch.
  - Fix incorrect variable name corresponding to DATA_EXPORT_ORDER
    making this directive inefficient. Thanks to Ron Johnson for the
    report.
  - Fix translation of check constraint when read from file
  - Fix EMPTY_LOB_NULL, enable/disable as inverted, keep default to
    enabled. Take care that in old ora2pg.conf it is disabled so it
    will break backward compatibility with old configuration.
  - Fix false positive detection of input filename is the same as
    output file.
  - Rename variables SCHEMA_ONLY, DATA_ONLY and CONSTRAINTS_ONLY in
    script import_all.sh to conform to their real use. Thanks to
    Sebastien Caunes for the report.
  - Fix comment detection breaking the package header parsing and
    global variable detection.
  - Fix ROWNUM detection for replacement by LIMIT
  - Fix escaping of psql command in configuration file comment and
    set default value for PG_VERSION to 12.
  - Replace precision by exactness in documentation. Thanks to
    Sebastien Caunes for the report.
  - Prevent reducing DATA_LIMIT when NO_BLOB_EXPORT is enabled.
   Thanks to Thomas Reiss for the report.
  - Fix geometry type detection.
  - Add autodetection of geometry type, srid and dimension for
    ArcGis geometries. Thanks to changmao01 for the feature request.
  - Fix call to ST_GeomFromText when no SRID is found.
  - Fix case where OVERRIDE SYSTEM VALUE clause could be added if PG
    version is < 10. Thanks to changmao01 for the report.
  - Fix unwanted call to internal GEOM library for ArcGis geometries.
    Thanks to changmao01 for the report.
  - Exclude schema SDE (ArGis) from export. Thanks to changmao01 for
    the report.
  - prevent looking twice to same custom data type definition.
  - Fix previous patch to catch SDO_GEOMETRY on lowercase regexp.
  - Limit detection of geometry data type to SDO_GEOMETRY.
  - Fix column name replacement in view definition. Thanks to Amit
    Sanghvi for the report.
  - Fix REPLACE_COLS parsing to allow space in column name. Thanks
    to Amit Sanghvi for the report.
  - Fix translation from file of triggers with WHEN clause. Thanks
    to Rui Pereira for the report.
  - Fix column name kept lowercase in the MOD() clause when -J is
    used. Thanks to Code-UV for the report.
  - Keep case of PG_SCHEMA definition when used in TEST action.
  - Fix data export for columns with custom data type. Thanks to
    Aymen Zaiter for the report.
  - Fix missing bracket with || operator in CREATE INDEX. Thanks to
    traxverlis for the report.
  - Fix export of single row unique function base index. Example:
      CREATE UNIQUE INDEX single_row_idx ON single_row ((1));
    Thanks to unrandom123 for the report.
  - Update documentation about schemas used in TEST action.
  - Disable materialized view export with MySQL export it is not
    supported. Thanks to naveenjul29 for the report.
  - Fix table alias detection in Oracle (+) join rewrite.
  - Fix an infinite loop in Oracle (+) join rewrite when there is no
    table aliases and the table is prefixed by its schema. Thanks to
    Olivier Picavet for the report.
  - Fix MODIFY_STRUCT when column name need to be escaped. Thanks to
    helmichamsi10 for the report.
  - Fix empty PARTITION BY () clause. Thanks to Aymen Zaiter.
  - Fix export of global variable from package description when there
    is no package body. Thanks to naveenjul29 for the report.
  - Add package description export when dumping package source,
    previously only the package body was dump. This will allow to
    check global variables export.
  - Whilst working on the Reproducible Builds effort (https//reproducible-builds.org/)
    it appears that ora2pg could not be built reproducibly. Thanks to
    Chris Lamb for the patch.
  - Fix case of NUMBER(*,10) declaration. Oracle has a precision of 1 to 38
    for numeric. Even if PostgreSQL allow a precision of 1000 use 38 to
    replace junk parameter. Thanks to xinjirufen for the report.
  - Add conversion of default value in function parameters, like syssdate
    rewriting for example. Thanks to unrandom123 for the report.
  - Fix a regression in data encoding when exporting data introduced in
    commit fa8e9de. Thanks to gp4git for the report.
  - Add debug information about the environment variables used before
    connecting to Oracle.
  - Fix case of duplicate between unique index and unique constraint with
    multiple columns. Thanks to gp4git.
Assets 2
Loading