# Changelog

## v0.9.0

- Support ERDDAP 2.10 search syntax (#155)

## v0.8.0

- Cache urlopen returns for faster access when the query does not change (#147)
- Major refactor that removes the utilities module (#145, #148)
- Build versioned docs with a drop-down menu to access previous versions (#146)
- Improved URL builder by avoiding duplicated slashes (#151)
- Normalize search terms to match ERDDAP server case (#152)

## v0.7.2

- Bugfix:

- Fixed Wrong type passed to load temporary netCDF files (#138)
- Fixed some f-strings bugs (#142)
- Allow erddapy to be loaded offline (#143)

## v0.7.1

- Bugfix: while griddap is not support the current ncCF enforced prevented users
          from building griddap URLs for xarray b/c griddap does not support it.
          Now the netCDF responses are ncCF for tabledap and plain nc for griddap.

## v0.7.0

- Added a ncCF response
- NetCDF file-like response can be in-memory or in-disk
  depending on the libnetcdf library available.
- Use canonical list of ERDDAP servers
- Various packaging improvements towards pep517/518
- Make `check_url_response` optional to avoid abusing servers
- New documentation with more examples and tips for multiple requests

Bugfix:

- Return ability to pass keyword arguments to `requests.get`.
- Fix `urlopen` `raise_for_status`
- Fix `parse_time_string` import for various pandas versions

## 0.6.0

- Added the `distinct()` server-side function
- Added type annotations and pre-commit checks as a GH-action
- Fetch ERDDAP servers list from a awesome-erddap

## 0.5.3

- Bugfix: fixed issue with pandas >=1.0.0.

## 0.5.2

 - Dummy release to trigger auto-publish

## 0.5.1

- Bugfix: fixed `get_search_url()` protocol filter in #94.

## 0.5.0

- Do not check server URL when creating the object.
  This allows servers that return 404 when the
  endpoint does not have any requests.

## 0.4.0

- Allow for multiple `csv`s responses in `.to_pandas`.

## 0.3.0

- Cache `_check_url_response` and `get_var_by_attr` returns
- Removed all python 2k code

## 0.2.0

- Add a `get_categorize_url` method
- Test suite for url_builder

## 0.1.11

- Re-factor methods that output a temporary netCDF file.

## 0.1.10

- Add a `to_iris` method.

## 0.1.9

- `erddapy` is not Python 3.7 compatible!
- Fixed a bug in the OPeNDAP URL endpoint.
- Use the `csvp` response in the `to_pandas()` method
  to merge the units in the column name.

## 0.1.8

- Fixed `setup.py` metadata.

## 0.1.7

- Fixed bug that preventing from generating download URLs with no data
  constraints or variables.

## 0.1.6

- Simplify the `erddapy` object constructor.
  Everything but the `server` and `protocol` should be passed as properties
  after the object creation.

## 0.1.5

- Re-implemented caching in `get_var_by_attr()`.

## 0.1.3

- Avoid a caching bug in `get_var_by_attr()` by not caching at all.

## 0.1.3

- New API to make it easier to handle multiple requests from the same server.

## 0.1.2

- Added a workaround to access authenticated ERDDAP servers.

## 0.1.1

- Mostly boilerplate updates for a GitHub/PyPI release.

## 0.1.0

- First version!
