- API urls
- are stored in
meta_info, a tibble created indata-raw/data_creation.R - This is where you can change from main to sandbox versions
- https://birdscanada.org/api vs. https://sandbox.birdscanada.org/api
- To apply this you must re-run
data-raw/data_creation.Rand then re-load the functions/package
- are stored in
Required and extra fields returned by the API are expected and treated, respectively, as follows:
| Function | API Entry Point | Required fields | Extra fields |
|---|---|---|---|
nc_count() |
data/list_collections |
collection, nrecords |
Ignored |
nc_count() |
data/list_permissions |
collection, akn_level |
Ignored |
nc_count() |
metadata/collections |
collection, akn_level |
Ignored |
nc_permissions() |
data/list_permissions |
collection, akn_level |
Ignored |
srv_auth() |
data/authenticate |
token |
Ignored |
nc_data_dl() |
data/release_request_id |
None (close request only) | Ignored |
nc_data_dl() |
data/get_data |
Any | Added |
meta_XXX() |
metadata/XXX |
Any | Added |
-
Missing "Required fields"
-
Missing/renamed "container"
Some API entry points return data in a container, for example,data/get_data->results(holds main data)data/list_collections->results(holds counts);request_id(holds request id)data/list_requests->requests(holds details on individual requests)
If these "containers" change names, the package will break.
- In package, after accessing the API, explictly
select()the fields/columns expected. This way extra fields won't break existing code- Missing fields will break the code, but at least they will break the code early!
- Do not do this for data downloads (i.e.
nc_single_dl()undernc_data_dl()) or metadata downloads (i.e.meta_XXX()) - Unless, using a
meta_XXX()download internally. Then alwaysselect()the fields required
- Tests are run using the user "testuser"
- Locally password can be stored in .Renviron as naturecounts_testuser = PASSWORD
- For remote testing, password is supplied as encrypted values