From 1ffa995a4a1c0ffe77fccfdc29c04777a946bd85 Mon Sep 17 00:00:00 2001
From: Colin Maudry <colin@maudry.com>
Date: Wed, 18 Mar 2026 22:02:20 +0100
Subject: [PATCH] docs: spec for distance histogram on observatoire, acheteur,
 titulaire pages

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---
 .../2026-03-18-distance-histogram-design.md   | 64 +++++++++++++++++++
 1 file changed, 64 insertions(+)
 create mode 100644 docs/superpowers/specs/2026-03-18-distance-histogram-design.md

diff --git a/docs/superpowers/specs/2026-03-18-distance-histogram-design.md b/docs/superpowers/specs/2026-03-18-distance-histogram-design.md
new file mode 100644
index 0000000..091e274
--- /dev/null
+++ b/docs/superpowers/specs/2026-03-18-distance-histogram-design.md
@@ -0,0 +1,64 @@
+# Distance Histogram — Design Spec
+
+**Date:** 2026-03-18
+**Branch:** feature/65_observatoire
+
+## Goal
+
+Display the distribution of distances (in km) between buyers and winning contractors, to help users assess whether a buyer or contractor tends to deal locally or at a national scale.
+
+## Data
+
+- Column: `titulaire_distance` (`Int64`, km)
+- Measured at address level — values are always > 0, no zero-handling needed
+- Already selected in the observatoire LazyFrame via `cs.starts_with("titulaire")`
+- Already available on acheteur and titulaire detail pages
+
+## Figure Function
+
+**Location:** `src/figures.py`
+
+**Signature:**
+
+```python
+def get_distance_histogram(lff: pl.LazyFrame) -> dcc.Graph:
+```
+
+**Behaviour:**
+
+- Collects `titulaire_distance` from the LazyFrame, drops nulls
+- If the resulting DataFrame is empty after dropping nulls, `px.histogram` produces a blank figure without errors — no guard logic needed. The order of operations must be: drop nulls → log-transform → histogram
+- Drop nulls first, then pre-log-transform the column (`pl.col("titulaire_distance").log(10)`) so bins are truly equal-width on a log scale. Use `px.histogram` with `nbins=50` on the transformed values
+- Set custom X-axis tick values at powers of 10 (1, 10, 100, 1000, 10000) with km labels, using `fig.update_xaxes(tickvals=[0,1,2,3,4], ticktext=["1","10","100","1 000","10 000"])`
+- Y axis: count of contracts
+- French axis labels: x = `"Distance (km)"`, y = `"Nombre de marchés"`
+- Returns a `dcc.Graph`
+
+## Integration
+
+### Observatoire (`src/pages/observatoire.py`)
+
+- `get_distance_histogram` imported and called inside `udpate_dashboard_cards`
+- Result wrapped in `make_card(title="Distance acheteur–titulaire", subtitle="en nombre de marchés, échelle logarithmique", fig=...)`
+- Card appended to the `cards` list alongside existing donuts and charts
+- No changes to the data pipeline — `titulaire_distance` is already in the LazyFrame
+
+### Acheteur page (`src/pages/acheteur.py`)
+
+The acheteur page uses a `dcc.Store` (`acheteur_data`) that holds serialised contract rows as a list of dicts. The integration follows the existing pattern used by other chart callbacks on this page:
+
+- Add a new `html.Div(id="acheteur-distance-histogram")` placeholder in the layout
+- Add a new callback with `Input("acheteur_data", "data")` that:
+  - Reconstructs `pl.LazyFrame(data)` from the store
+  - Calls `get_distance_histogram(lff)`
+  - Wraps the result in `make_card(...)` and returns it to the placeholder div
+
+### Titulaire page (`src/pages/titulaire.py`)
+
+Same pattern as acheteur: `dcc.Store` (`titulaire_data`) → new callback → `html.Div` placeholder.
+
+## Out of Scope
+
+- Filtering by distance range (could be a future filter on the observatoire page)
+- Showing distance on a map or as a trend over time
+- Bucket-based (named zone) grouping