As a user, when I declare a baseline dataset with a different variable name than the predicted dataset, I expect success

[james-d-brown]

Given an evaluation that contains a different variable name for each of the predicted and baseline datasets When the retrieval queries are formed Then I expect the appropriate name to be used for each dataset and not the same name for both

[james-d-brown]

From a Redmine user support ticket, #135539.

[james-d-brown]

To reproduce, use this declaration:

label: HEFS_Benchmark Streamflow
observed:
  label: OBS Streamflow
  sources: /data/BONM8_QME.xml
  variable: QME
  feature_authority: nws lid
  type: observations
predicted:
  label: HEFS_Benchmark Streamflow
  sources: /data/200101011200_BONM8_RAW_QINE.xml
  variable: QINE
  type: ensemble forecasts
baseline:
  label: HEFS_Raw Streamflow
  sources: /data/200101011200_BONM8_RAW_SQIN.xml
  variable: SQIN
  type: ensemble forecasts
  separate_metrics: true
features:
  - {observed: BONM8, predicted: BONM8, baseline: BONM8}
unit: cms
lead_times:
  minimum: 18
  maximum: 720
  unit: hours
lead_time_pools:
  period: 24
  frequency: 24
  unit: hours
time_scale:
  function: mean
  period: 24
  unit: hours
pair_frequency:
  period: 24
  unit: hours
#cross_pair: exact
probability_thresholds: [0.01,0.1,0.5,0.9,0.95,0.99,0.995,0.999]
minimum_sample_size: 30
season:
  minimum_day: 1
  minimum_month: 1
  maximum_day: 31
  maximum_month: 12
metrics:
  - box plot of errors by forecast value
  - continuous ranked probability score
  - relative operating characteristic score
  - brier score
  - pearson correlation coefficient
  - root mean square error
  - mean error
  - bias fraction
  - reliability diagram
  - relative operating characteristic diagram
  - ensemble quantile quantile diagram
  - brier skill score
  - name: quantile quantile diagram
    thresholds: all data
  - continuous ranked probability skill score
  - sample size
  - mean absolute error
  - box plot of errors by observed value
  - name: rank histogram
    probability_thresholds:
      values: [0.01,0.1,0.5,0.9,0.95,0.99,0.995,0.999]
      apply_to: predicted
duration_format: hours
decimal_format: '#0.000'
output_formats:
  - csv2
  - format: png
    orientation: lead threshold

Together with the attached datasets (unpack first):

data.zip

Witness a failure to create pairs for any baseline dataset.

The underlying queries for baseline time-series look like this:

SELECT
    metadata.series_id AS series_id,
    metadata.reference_time + INTERVAL '1' MINUTE * TSV.lead AS valid_time,
    metadata.reference_time,
    metadata.reference_time_type,
    ARRAY_AGG(TSV.series_value ORDER BY TS.ensemble_id) AS ensemble_members,
    ARRAY_AGG(TS.ensemble_id ORDER BY TS.ensemble_id) AS ensemble_ids,
    metadata.measurementunit_id,
    metadata.scale_period,
    metadata.scale_function,
    metadata.feature_id,
    metadata.occurrences
FROM
(
    SELECT
        S.source_id AS series_id,
        MAX( reference_time ) AS reference_time,
        TSRT.reference_time_type,
        S.feature_id,
        S.measurementunit_id,
        TimeScale.duration_ms AS scale_period,
        TimeScale.function_name AS scale_function,
        COUNT(*) AS occurrences
    FROM wres.Source S
    INNER JOIN wres.ProjectSource PS
        ON PS.source_id = S.source_id
    INNER JOIN wres.TimeSeriesReferenceTime TSRT
        ON TSRT.source_id = S.source_id
    LEFT JOIN wres.TimeScale TimeScale
        ON TimeScale.timescale_id = S.timescale_id
WHERE PS.project_id = 43
    AND S.variable_name = 'QINE'
    AND S.feature_id = 1556
    AND PS.member = 'baseline'
    GROUP BY S.source_id,
        S.measurementunit_id,
        TimeScale.duration_ms,
        TimeScale.function_name,
        TSRT.reference_time_type
) AS metadata
INNER JOIN wres.TimeSeries TS
    ON TS.source_id = metadata.series_id
INNER JOIN wres.TimeSeriesValue TSV
    ON TSV.timeseries_id = TS.timeseries_id
WHERE TSV.lead > 39960
    AND TSV.lead <= 42840
    AND ( EXTRACT( MONTH FROM metadata.reference_time ) > 1 OR ( EXTRACT( MONTH FROM metadata.reference_time ) = 1 AND EXTRACT( DAY FROM metadata.reference_time ) >= 1 ) )
    AND ( EXTRACT( MONTH FROM metadata.reference_time ) < 12 OR ( EXTRACT( MONTH FROM metadata.reference_time ) = 12 AND EXTRACT( DAY FROM metadata.reference_time ) <= 31 ) )
GROUP BY metadata.series_id,metadata.reference_time, metadata.reference_time_type, metadata.feature_id, TSV.lead, metadata.scale_period, metadata.scale_function, metadata.measurementunit_id,metadata.occurrences

Crucially:

AND S.variable_name = 'QINE'

Contrary to the declaration:

baseline:
  label: HEFS_Raw Streamflow
  sources: /data/200101011200_BONM8_RAW_SQIN.xml
  variable: SQIN