Summarize per-pulse telemetry

Usage

ledgr_backtest_bench(bt)

Arguments

bt: A ledgr_backtest object. This function does not accept a DuckDB file path; use ledgr_run_info() for persisted run-level telemetry.

Value

A tibble with mean/median/p99 values per telemetry component.

Details

This is a diagnostic helper for engine profiling. It only reports detailed telemetry captured for runs executed in the current R session. Timing components are reported in seconds; feature-cache hit/miss rows are counts. The compact telemetry persisted in durable experiment stores is available through ledgr_run_info().

Examples

bars <- data.frame(
  ts_utc = as.POSIXct("2020-01-01", tz = "UTC") + 86400 * 0:2,
  instrument_id = "AAA",
  open = c(100, 101, 102),
  high = c(101, 102, 103),
  low = c(99, 100, 101),
  close = c(100, 101, 102),
  volume = 1000
)
strategy <- function(ctx, params) {
  targets <- ctx$flat()
  targets["AAA"] <- 1
  targets
}
bt <- ledgr_backtest(data = bars, strategy = strategy, initial_cash = 1000, cost_model = ledgr_cost_zero())
ledgr_backtest_bench(bt)
#> # A tibble: 15 × 4
#>    component                mean   median      p99
#>    <chr>                   <dbl>    <dbl>    <dbl>
#>  1 t_pre                 1.13     1.13     1.13   
#>  2 t_post                0.0120   0.0120   0.0120 
#>  3 t_loop                0.01000  0.01000  0.01000
#>  4 t_pulse              NA       NA       NA      
#>  5 t_bars               NA       NA       NA      
#>  6 t_ctx                NA       NA       NA      
#>  7 t_fill               NA       NA       NA      
#>  8 t_state              NA       NA       NA      
#>  9 t_feats              NA       NA       NA      
#> 10 t_strat              NA       NA       NA      
#> 11 t_target             NA       NA       NA      
#> 12 t_event              NA       NA       NA      
#> 13 t_exec               NA       NA       NA      
#> 14 feature_cache_hits    0        0        0      
#> 15 feature_cache_misses  0        0        0      
close(bt)