Debugging

log.info("entering, close=" + str.tostring(close))
log.info("rsi={0,number,#.##} atr={1}", ta.rsi(close, 14), ta.atr(14))   // str.format-style placeholders

if ta.crossover(close, ta.sma(close, 20))
    log.warning("Cross up on " + str.tostring(bar_index))

if close < 0
    log.error("Negative close — bad data?")

• One entry is written each time the call executes: once per historical bar, and on every tick of the live bar.
• Each entry carries the level, the bar time (the bar's open time; the system time is shown on hover), the script (id@version), the message and the source line of the log.* call.
• Live bar: the entries of the open bar are replaced on every tick (the pane always shows the latest tick's lines for that bar); once the bar closes they are final. They are not rolled back the way plots are.
• Caps: at most 500 lines per bar (a warning log limit reached — further entries dropped is appended once) and 10,000 per execution; the Logs pane keeps the latest 10,000. Logs survive until the script restarts (re-publish, inputs change, symbol or timeframe change).
• log.* is not allowed inside a request.security expression (it runs in another context).

Open it with the Logs button in the editor header. It merges the logs of every execution you currently have on charts and offers:

Tips: a log.* call inside a loop multiplies per bar and eats the 500-line budget quickly — collect loop values into an array or a string and log once after the loop; gate noisy calls behind an input.bool or a time range (input.time) so the useful history reaches further back.

The oldest trick: plot a numeric value you can read off the chart.

raw = (close - open) / open * 100
plot(raw, "body %", display=display.data_window)

display=display.data_window keeps the value out of the pane but puts it in the Data Window read-out so you can hover individual bars.

A tiny table in a corner is often the fastest way to see the running state of several variables:

var table t = table.new(position.bottom_right, 2, 4, bgcolor=color.new(color.black, 20))
if barstate.islast
    table.cell(t, 0, 0, "RSI",     text_color=color.gray,  text_halign=text.align_left)
    table.cell(t, 1, 0, str.tostring(ta.rsi(close, 14), "0.00"),
                                  text_color=color.white, text_halign=text.align_right)
    table.cell(t, 0, 1, "ATR",     text_color=color.gray,  text_halign=text.align_left)
    table.cell(t, 1, 1, str.tostring(ta.atr(14), "0.00"),
                                  text_color=color.white, text_halign=text.align_right)

A log.* call before and after a suspected hot section narrows a hang:

log.info("before heavy loop")
for i = 0 to 1_000_000 by 100
    …
log.info("after")

The last message that fires localises the problem.

The editor's Problems list shows compile diagnostics as you type. Click an entry to jump to the offending line. See Errors overview.

• A compile error prevents publish — you'll see it in the editor before the script runs.
• A runtime error (runtime.error, a budget overrun, a request without data) stops the execution and shows on the indicator's legend and in the Problems strip; warnings injected by the platform (no data for a symbol, log limit reached) land in Logs and the script keeps running.

When the question is how slow rather than what value, switch on Profiler mode (Script menu): every significant line gets its share of runtime and its execution count right beside the code. See Profiling and optimization.

• Errors overview
• Profiling and optimization