Minor bugfix release

• #435 Fixed 'INDEXED BY' clause followed by multiple identifiers.
• #437 Added support for '@' characters inside identifier names.

Minor feature release

• #430 Added support for 'twisted' RDW headers when big-endian or little-endian RDWs use unexpected RDW bytes.

Minor feature release

• #420 Add experimental support for fixed blocked (FB) record format.
• #422 Fixed decoding of 'broken pipe' (¦) character from EBCDIC.
• #424 Fixed an ASCII reader corner case.

Feature Release

• #412 Add support for variable block (VB aka VBVR) record format.
  Options to adjust BDW settings are added:
  • is_bdw_big_endian - specifies if BDW is big-endian (false by default)
  • bdw_adjustment - Specifies how the value of a BDW is different from the block payload. For example, if the side in BDW headers includes BDW record itself, use .option("bdw_adjustment", "-4").
  • Options is_record_sequence and is_xcom are deprecated. Use .option("record_format", "V") instead.
• #417 Multisegment ASCII text files have now direct support using record_format = D.

Feature Release

• #405 Fix extracting records that contain redefines of the top level GROUPs.
• #406 Use 'collapse_root' retention policy by default. This is the breaking,
  change, to restore the original behavior add .option("schema_retention_policy", "keep_original").
• #407 The layout positions summary generated by the parser now contains level
  numbers for root level GROUPs. This is a breaking change if you have unit tests that depend on the formatting of the layout
  positions output.

Minor feature release

• #397 Fix skipping of empty lines when reading ASCII files with is_record_sequence = true
• #394 Added an ability to specify multiple paths to read data from (Use .option("paths", inputPaths.mkString(","))). This is a workaround implementation since adding support for multiple paths in load() would require a big rewrite for spark-cobol from data source to data format.
• #372 Added an option to better handle null values in DISPLAY formatted data: .option("improved_null_detection", "false")