Libhxl

- add __version__ attribute to module
- refactor the hxl.validation module for better testing and
maintainability
- add a new default schema with useful default tests
- allow multiple tag patterns (comma-separated) for valid_tag in
a HXL schema
- add a spelling validation test
- add a numeric-outlier validation test
- refactor CacheFilter to preserve row numbers
- add ability to generate a JSON-style validation report easily
via hxl.validate()
- new requirement: python-io-wrapper
- RowFilter (with_rows, without_rows) no longer ignores empty
cells; that will occasionally give some different results
- when multiple columns match a row query, it will succeed with at
least one success
- fixed a bug parsing "is" row queries
- handle more Google Sheets URLs
- recognise datetime formats as dates

- hotfix for bug in date parsing

- remove obsolete Python2 compatibility code
- added source_row_number and source_column_number to support validation
- add wildcard support to tag patterns, so that we can use
patterns like "*" or "*+f-children"
- revamped date handling to support partial dates like "2018-01"
or "2018", and also special notation like "2018W05" or "2018Q1"
- add min and max methods to hxl.model.Dataset
- HXL validation reports a validation error when a valid_value+url is not usable
- HXL validation now reports proper column
- HXL validation now accepts all parseable date formats
- HXL validation now has a valid_unique constraint (single value or
compound key)
- HXL validation now has a valid_correlation constraint (e.g. make sure
that adm1 and adm2 are always consistent for any given value
of adm3)
- HXL validation can now try to infer datatypes without explicit rules
- HXL validation now calculates edit distance and suggests the
closest match when failing validation against a list
- HXL validation can now test for irregular whitespace using
valid_value+whitespace
- add "is (not) min" and "is (not) max" support to hxl.model.RowQuery
- add is_cached flag to hxl.model.Dataset and subclasses
- updated all AbstractInput to be iterables rather than iterators
(for repeatability)
- removed hxl.common module and added hxl.datatypes, with
more-consistent data checking/conversion
- when importing JSON arrays and objects, flatten them to a usable
text representation
- update docstrings
- default to case insensitive for validation
- added static hxl.model.TagPattern.match_list method
- fixed hxl.filters.ReplaceDataFilter to allow multiple tag
patterns
- fixed bug when an empty row appears before the hashtag row

- end support for Python 2 (will die with a RuntimeError;
next release will remove Python2 compatibility code relics)
- start implementing logging support
- all command-line utilities now have a --log option to set the
logging level
- restore support for preserving original attribute order (except
for JSON object flavoured export)
- add hxl.Column.get_display_tag method with optional attribute
sorting
- add support for lat/lon normalisation to clean-data filter
- add purge option to clean_data to allow removing numbers, dates,
or lat/lon that can't be parsed during data cleaning
- fix bug opening a Google Sheet from a CKAN resource URL
- allow opening a dataset from a CKAN dataset URL (uses first
resource)
- make the 'patterns' parameter optional for the JSON count recipe

- bug-fix release: do not let a misspelled date cause a fatal
exception

- the merge-data filter now looks for keys in *all* candidate
columns (not just the first-matching ones)
- add skip_untagged parameter to without_columns and ColumnFilter,
for removing columns without HXL hashtags
- hxl.model.Row.get_all can take a default value
- the clean-data filter has a number_format option (e.g. "0.2f")
- the hxlclean command-line script has a --number-format option
- hxl.model.Column.display_tag always shows attributes sorted, per HXL
1.1 beta
- added hxl.model.Row.dictionary property to return row as a
Python dict
- hxl.model.Source.gen_json() has a new use_objects option to use
the JSON list-of-objects format from HXL 1.1 beta as output
- hxl.io.write_json() has the use_objects option to pass on to
hxl.model.Source.gen_json()
- try to recognise JSON data even if it doesn't have a JSON MIME
type or file extension