POSSIBLE BUG - discrepancy between prod summary table data and expected from manual calculations

[mwalker-scottlogic]

Observed variation between actual summary table data for London (forecast base time: 18 Aug 12:00, forecast window: 1), and expected data from a manual calculation.

Steps to reproduce:

• query in_situ_data database for:

{ $and: [
  {"measurement_date": { 
    $gte: new ISODate('2024-08-18T10:30:00.000+00:00'), 
    $lte: new ISODate('2024-08-19T13:30:00.000+00:00')}
  },
  {"name": "London"}
]}

• at each 3hr increment, for each pollutant, average the data for the surrounding +/- 1.5hrs

• calculate AQI

• query forecast_data for:

{ $and:[
  {"forecast_base_time": new ISODate('2024-08-18T12:00:00.000+00:00')},
  {"name": "London"}
]}

• extract from forecast data the first 9 valid times as they will correspond to the time range needed
• compare UI table for this time range with the data collected

Expected:

• data in UI corresponds to values that are the greatest difference between forecast and measured AQI
• the values in the table are the same

Actual:

• some variance discovered up to 3.15%

• Some unpredictable behaviour in terms of which time stamp is selected by UI to represent the diff, sometimes it is first time, sometimes a different one selected

Test Checklist:

• Test Analysis & discussion with developer
• Write Test Cases / Charters
• Test Cases & Charters Reviewed
• Test Cases Executed
• Test Charters Executed
• Regression tests executed
• Automation Tests added to project

[mwalker-scottlogic]

Test Analysis

Having discussed the issue with @rstrange-scottlogic, it was actually an issue in my calculation that found the divergence - excel was treating long numbers as text and excluding them from the mean calculation.

A change has been made where previously if the measured AQI and forecasted AQI were both 1 for the entire duration of the time, the UI picked the last entry, it now picks the first

• No automation required.
• Need to re-execute the manual test to verify values ✅
• Need to check if the UI uses the first time slot in the above scenario ✅

Charter Re‐execute: Charter 28: Explore production summary table with reference to the database to discover accuracy ✅

[mwalker-scottlogic]

All that is left is for this to undertake regression testing

[mwalker-scottlogic]

Evidence from re-running charter that the UI is picking the first time slot when the max diff is shared

[mwalker-scottlogic]

Frontend regression PASSED ✅

Playwright Regression 29.08.24.pdf

Backend regression PASSED ✅