01The gap this fills
Everything numeric already existed. Performance.Metrics computes 27 risk-adjusted figures, Drawdown_analysis extracts episodes with recovery times, Attribution decomposes P&L, Risk_management.Var prices four VaR methods.
What was missing was any way to get them out. Every one of those layers exposes a to_string aimed at a terminal — 48 such functions across the tree — and not one exposes rows. There was no to_json anywhere, and the only CSV writers were two hand-rolled Printf loops in Backtest.Result.
02A real bug it fixes
Backtest.Result.write_blotter_csv emitted strategy_id and tag with a bare %s. Both are strategy-supplied free text — tag exists precisely so a fill can be traced back to the reason it was placed — so a comma or a newline in either silently corrupted the file, shifting every later column by one.
Export.csv_escape implements RFC 4180 quoting: a field containing a comma, a double quote, CR or LF is wrapped in quotes with internal quotes doubled. Result's writers now use the same rule.
Non-finite floats render as an empty cell, not nan: a spreadsheet treats nan as text and it poisons the whole column's type inference.
03Reports
| Name | Contents |
|---|---|
performance | Every metric from a NAV curve, via Metrics.to_assoc |
drawdowns | Episodes: peak, trough, recovery, depth, duration |
risk | Exposure, VaR/ES, limit breaches, circuit-breaker state |
positions | Everything currently held, per instance |
fills | Execution view: price, commission, maker/taker |
audit | The audit trail — see below |
summary | The runtime's flattened metrics |
Fetch any of them as CSV or JSON:
curl 'localhost:8080/api/reports/fills?format=csv'04Not xlsx
Excel export is not implemented. A real .xlsx is a ZIP container of OOXML parts, which is a meaningful amount of machinery for a format Excel opens from CSV anyway. Requesting it returns a 400 that says so, rather than silently handing back CSV under a misleading name.
05"Regulatory reporting"
A stated goal was "regulatory reporting capabilities". What ships is audit: a complete, timestamped record of every simulated order and fill, carrying the identifiers needed to reconcile an order end to end, plus an explicit execution_mode column reading paper.
It is not certified against MiFID II RTS 22, CAT, EMIR or any other regime, and nothing here should be filed with anyone. Calling an audit export "regulatory reporting" without that sentence would be the kind of claim this project takes care not to make.
06Known gaps / follow-ups
- Portfolio-level performance needs a NAV curve the runtime does not retain. Each instance keeps its own bounded ring; there is no combined, durable series, so
performanceanddrawdownsover the live runtime report on an empty curve rather than inventing data. Run them against a backtestResultinstead. - No scheduled generation. Reports are pulled; nothing writes them on a timer.
- No attribution endpoint.
Report.attributionexists and works, but the daemon has no fills array in the shapeAttributionwants without a conversion pass.
07Source map
| Module | Path |
|---|---|
| CSV/JSON writers, quoting | lib/reporting/export.{ml,mli} |
| Report assembly | lib/reporting/report.{ml,mli} |
| Escaping fix in the backtest writer | lib/backtest/result.ml |
| Tests | test/reporting/test_export.ml |
08Scope and limitations
Performance, risk and execution reports generate and export as CSV and JSON.
Regulatory reporting ships as an uncertified audit trail — a defensible record of activity, not a compliance claim against any specific regime. .xlsx export is not implemented; CSV opens natively in Excel.