Libhxl

- throw proper exception for failed HTTP request from requests library
- support JSON arrays of objects as well as arrays of arrays
- recursively search for HXL data inside a JSON dataset

- fixed bug with disabling SSL checks via requests library
- improved Excel data handling — use integers instead of floats
when possible, and fix bug when trying to parse numbers as dates

- support JSON input (list of rows)
- add fill_data filter to fill empty cells from previous rows
- use MIME type and extension where available to help choose type
- grab character encoding from HTTP response if available
- add verify_ssl parameter to hxl.io.data, hxl.io.make_input,
etc. Defaults to True; if False, don't try to validate SSL certs
- add new "is (not)" operator for row queries
- add optional date_format parameter for clean data filter
- fix bug with error messages from scripts
- fix output bug in command-line scripts
- fix bug in merge filter
- fix bug in hxl.io.tagger

- Fully modularised JSON specs and recipes.
- Made JSON specs and recipes work recursively.
- Fix bug that caused select filter to fail after explode filter
- Added top-level hxl.tagger() function, similar to hxl.data()
- add optional default_header arg to hxl.model.Column.parse_spec
- major overhaul of the merge_data filter: now merges *all* columns
matching the pattern supplied, and doesn't create an empty
column if there are no matching columns in the merge dataset
- refactored append filter to allow multiple append files in
single filter

- Regex ~ and !~ operators in row queries now match anywhere in the cell
- make row query smart about date comparisons with date hashtag
- block numeric/date conversion in row queries for ~ and !~

- add timeout option for opening URLs (avoids long wait in unit
tests)
- encoding fixes for Python2
- add HXLColumn.has_attribute() method
- add an optional parsed attribute to HXLRow.get() to try parsing
the value according to attributes (currently supports +list)
- add experimental support for the +list attribute