| |
|
|
|
|
| |
With the default file scope, the dashboard rendered and exported from its own buffer, so every count read zero and cards.csv came out empty. The render, refresh, cycle, and CSV-export paths now run their card scans in the deck buffer the command was invoked from, stored buffer-locally, and a dead deck signals a clear user-error instead of scanning the wrong buffer.
Card links are now a registered org-drill-card link type with a follow handler, and the link path carries the file identity so RET jumps work across files. Eleven new ERT tests cover the render, refresh, export, link, and dead-buffer paths.
|
|
|
Step 1 shipped the session-log data layer. This is the renderer on top of it.
org-drill-statistics opens a read-only dashboard with five sections: an overview (card counts plus a last-session recap), trends (reviews-per-day and pass-rate-per-day quadrant-block sparklines over the trend window, plus a 12-week table), a quality histogram, a needs-attention view (leech candidates, long-overdue, and forgotten-new cards), and a 7-day forecast counted from SCHEDULED dates. A buffer-wide filter (scope, range, algorithm) sits in the header and cycles with s/r/a. The other keys are q to bury, g to refresh, e for the CSV-export hook that lands next, and RET to follow the card link at point.
The aggregation math lives in pure helpers (day-bucketing, sparkline scaling, weekly aggregates, the histogram, the attention selectors, forecast bucketing). The render helpers are thin string formatters over them, so the logic is unit-tested independently of the UI. New defcustoms tune the views: org-drill-statistics-trend-days, -forecast-days, -attention-row-limit, and -leech-quality-threshold.
I added require 'calendar for the Monday week-start arithmetic in the weekly aggregates. CSV export and the manual and README entries are the step-3 follow-on.
|
