[Mehulsri07]

Phase 3 of the design: the first screens that use the model.

What it adds

Map: pick any spot

• Clicking anywhere that is not a borewell marks the spot with a dashed ring and shows Proposed layers here in the side panel.
• The panel says how many borewells the proposal comes from and how far the nearest is, and draws the column.
• The column never looks measured: faded with dashed edges where it is likely, paler where it is a rough guess, and left empty and hatched where there is not enough data.
• Clicking a stretch of the column says what it is made of in words ("Mostly water-bearing sand, some clay") and lists the borewells behind it with their distances.
• Clicking a borewell pin shows its own recorded layers, solid, in their own colours.
• Far from every borewell, or outside the ground-height data, it says there are not enough borewells to say anything.

Data quality

• Plain sentences from the accuracy check: how often "Likely" and "Rough guess" were right, and how that changes with distance. A sentence is left out when it would rest on too few steps to mean anything.
• A list of borewells that do not fit their surroundings, to look at in StrataField.
• With fewer than 10 usable borewells it says it cannot check yet.

Not done here, on purpose

• The confidence rules are still the starting values. Tuning them needs the real records, which are not on the development computer. The Data quality screen is the tool for it: run this build where the real records are and read what it says.
• The accuracy check runs on the main thread, not in the background as the design said. It takes well under a second at 500 borewells; there is a note in the code on when to move it.

Checks

• 102 TypeScript tests, including the wording functions and the water-level rules.
• The real-app test now has a second scenario: started on made-up borewells, it clicks a spot, expects proposed layers and a water level, opens Data quality and expects the one borewell that was logged upside down.
• Tried by hand in the real app on made-up borewells, with screenshots looked at: a spot between borewells, a click on a layer, a click on a borewell, a spot far from everything, and Data quality.

Corrected after review

• Water level. Only dated readings from the last few years count, only within 1 km (the same reach StrataField's water map uses), and the panel says how far the nearest reading is. Undated readings were being treated as recent. The expected level is drawn dashed; a borewell's measured level is solid and says the year.
• A selected stretch is described by the average of everything in it and lists every borewell behind it. Before, it was described by its middle 5 ft alone.
• Borewells at one spot. Only the top pin can be clicked, so the panel now lists all the borewells at that spot to choose from.
• The map is measured again when it comes back from behind Data quality; it could have come back blank after a window resize.
• Data quality needs at least 5 borewells behind a sentence, not only 50 steps, and describes odd ones in feet.