NEWS.md
issued_at
argument to upload_truth()
upload_forecast()
function to accept a new is_json
argument, making the new signature: upload_forecast(zoltar_connection, model_url, timezero_date, forecast_data, is_json = TRUE, notes = "")
create_timezero <- function(zoltar_connection, project_url, timezero_date, data_version_date, is_season_start, season_name)
functionjob_info(zoltar_connection, job_url)
to support add missing timezero, unit, and target info to truth upload’s job.output_json #344 - by parsing job_json$output_json$missing_time_zeros
into date objects.outcome_variable
to all targets, and to add numeric_horizon
and reference_date_type
fields to is_step_ahead
ones. These changes support upcoming visualization features.truth_info()
to match a server change in truth naming: now has source
and created_at
instead of truth_csv_filename
and truth_updated_at
, respectivelyforecast_info()
and forecasts()
to support forecast issue_datejob_data()
(and therefore do_zoltar_query()
) to parse the quantile column as double (was defaulting to logical in some cases)models()
: added “notes” columnforecasts()
: added three timezero columns in addition to “url”: “timezero_date”, “data_version_date”, “is_season_start”scores()
function (removed from the library) that downloaded all scores to querying scores in a manner very similar to that of forecasts. This entailed generalizing do_zoltar_query
, submit_query()
, and job_data()
, which now take a is_forecast_query
arg. The first two now accept either "types"
or "scores"
, depending on that arg. 2) Changed those two score query functions to pass strings, not database IDs, removing the need to call query_with_ids()
, which has been deleted.do_zoltar_query()
query args to all be optional.job_data()
to return units
column as character, not numeric.units
arg to do_zoltar_query()
to match submit_query()
.busy_poll_job()
to check for the new TIMEOUT status.query_with_ids()
’s query
arg to accept model abbreviations in “models”, instead of passing them in “model_abbrs”, which is no longer accepted. This is because Zoltar now uses model abbreviations to uniquely identify them, not names.do_zoltar_query()
and busy_poll_job()
.