Repository navigation
Four valid request.security forms refused: lookahead from a variable, '' timeframe, [] on a requested value in a nested request, an uncalled wrapper #294
Description
Summary
Each script below is valid v6 and is refused at translation. Each has a control that differs in one place and translates and runs.
| # | Pine form | Refusal |
|---|---|---|
| 1 | lookahead = mode, with mode chosen from an input.bool |
ANALYZER request.security lookahead must be barmerge.lookahead_on or barmerge.lookahead_off. |
| 2 | timeframe = '' |
ANALYZER request.security: invalid timeframe literal ''. |
| 3 | ta.ema(base[1], 5) inside a second request, where base is the result of a first request |
CODEGEN request.security helper call history is only supported in the payload itself |
| 4 | a function that wraps request.security and is never called |
CODEGEN Unknown variable 'expr' |
Versions
pineforge-codegen 1.9.1, and main at c1cc532 (after #290) run from source: same refusals. Controls were built with pineforge-engine v1.9.1 (fc0a2df), Apple clang 17.0.0, macOS 15.8 arm64, and run on BTCUSDT 1h bars from 2024-01-01 to 2024-03-01. The refusals come from the front end, so they should not depend on the platform.
Reproduction
1. lookahead from a variable
//@version=6
strategy('repro')
useLookahead = input.bool(true, 'Lookahead')
mode = useLookahead ? barmerge.lookahead_on : barmerge.lookahead_off
daily = request.security(syminfo.tickerid, 'D', close[1], lookahead = mode)
plot(daily)
ANALYZER 5:71 request.security lookahead must be barmerge.lookahead_on or barmerge.lookahead_off.
Control, lookahead = barmerge.lookahead_on: translates and runs.
2. Empty timeframe
//@version=6
strategy('repro')
same = request.security(syminfo.tickerid, '', close)
diff = same - close
// @pf-trace diff=diff
plot(same)
ANALYZER 3:43 request.security: invalid timeframe literal ''. Expected Pine TF format like '1', '15', '1H', '1D', '15S'.
Control, timeframe.period in place of '': translates, and diff is 0 on all 1,441 bars.
3. [] on a requested value inside a second request
//@version=6
strategy('repro')
base = request.security(syminfo.tickerid, 'D', close)
smooth = request.security(syminfo.tickerid, '240', ta.ema(base[1], 5))
plot(smooth)
CODEGEN 1:1 request.security helper call history is only supported in the payload itself
Control, ta.ema(base, 5): translates and runs. The refusal points at line 1, column 1, not at the call. This is a different form from #287: no user function is involved.
4. A wrapper function that is never called
//@version=6
strategy('repro')
fetch(sym, tf, expr) => request.security(sym, tf, expr)
daily = request.security(syminfo.tickerid, 'D', close[1])
plot(daily)
CODEGEN 3:51 Unknown variable 'expr' — not a PineForge builtin or a declared variable.
Control, without the fetch line: translates and runs. When the wrapper is called, daily = fetch(syminfo.tickerid, 'D', close[1]), the script translates and runs, so only the uncalled definition fails.
Expected
All four translate and give the controls' values.
lookaheadtakes a "simple" value, and a ternary of an input is "input", which is weaker and accepted (qualifiers,lookahead). Checked on TradingView (BINANCE:BTCUSDT 1h): the script compiles and runs, with no error or warning.- An empty string as the timeframe means the chart's timeframe, the same as
timeframe.period(timeframes). Checked on TradingView (BINANCE:BTCUSDT 1h): the script compiles and runs;diffis 0 on every bar. - A request whose expression depends on another request evaluates that request in its own context (nested requests), and
[]applies to any series (history-referencing operator). If the form cannot be supported, a refusal located at the call would help. Checked on TradingView (BINANCE:BTCUSDT 1h): the script compiles and runs, with no error or warning. - A function that is never called does not affect the script (user-defined functions). Checked on TradingView (BINANCE:BTCUSDT 1h): the script compiles and runs, with no error or warning, and
dailyhas the same values as in the control.
Workaround
- Write two
request.securitycalls with literallookaheadvalues and pick the result with the input:daily = useLookahead ? dailyOn : dailyOfftranslates and gives the same values as the control. - Use
timeframe.period. - None found that keeps the same values.
- Remove the unused wrapper.
Activity
luisleo526 commentedon Nov 1, 2026
Thank you for the detailed report. We are checking each case against TradingView's behaviour and our source, and will post a result for each case in this issue by the end of 2026-11-03 (UTC). If a case needs longer, we will say so here with the reason.
luisleo526 commentedon Nov 3, 2026
Thank you for the four scripts, their controls and the TradingView checks. This is the result for each case on the released pineforge-codegen 1.10.0. None of the four is fixed in 1.10.0; the fix is planned for the next release (1.10.1), as the 1.10.0 release notes list for this issue.
lookahead = mode, withmodechosen from aninput.boolrequest.security lookahead must be barmerge.lookahead_on or barmerge.lookahead_off.timeframe = ''request.security: invalid timeframe literal ''.''means the chart's timeframe on TradingView, astimeframe.perioddoesta.ema(base[1], 5)inside a second request, wherebaseis the result of a first requestrequest.security helper call history is only supported in the payload itself(reported at line 1, column 1, not at the call)request.securityand is never calledUnknown variable 'expr'; when the wrapper is called the script translates and runsAll four are defects, not intended limits. The fix is planned for the next release (1.10.1); the issue stays open until it is released. Until then your workarounds hold (two requests with literal
lookaheadvalues and a selection;timeframe.period; removing the unused wrapper); for case 3 no workaround that keeps the values has been found.luisleo526 commentedon Nov 6, 2026
Status: all four cases are fixed on main: each script now translates, builds and runs. The fixes ship in the next release; the issue stays open until they are released.
Checked with the code of main at 5b72a0c and the engine at d398279, on Linux aarch64: each script of the issue was translated, the generated C++ was built against the engine and run on BTCUSDT 1h bars from 2024-01-01 to 2024-03-01 (1,441 bars). "Result on 1.10.0" is the result in our comment of 2026-11-03. The TradingView results are the ones the report gives.
lookahead = mode, withmodechosen from aninput.boolrequest.security lookahead must be barmerge.lookahead_on or barmerge.lookahead_off.dailyequals the control's (lookahead = barmerge.lookahead_on, the default input) on all 1,441 barstimeframe = ''diffis 0 on every barrequest.security: invalid timeframe literal ''.diffis 0 on all 1,441 barsta.ema(base[1], 5)inside a second request, wherebaseis the result of a first requestrequest.security helper call history is only supported in the payload itselfsmoothhas its first value on bar 27request.securityand is never calleddailyhas the control's valuesUnknown variable 'expr'dailyequals the control's (without the wrapper) on all 1,441 barsNot run: x86 and macOS.
luisleo526 commentedon Nov 8, 2026
Update: today's check confirms that all four reported forms translate, build and run on main.
Checked on 2026-11-08 on codegen main at
7d7681dcwith engine main at755e6658, on the 1,441-bar BTCUSDT 1h sample.lookahead = mode, withmodechosen from aninput.booldailyequals thelookahead_oncontrol on all 1,441 bars with the default input08261a24)timeframe = ''diffis 0 on all 1,441 bars08261a24)ta.ema(base[1], 5)inside a second request, wherebaseis the result of a first requestsmoothfirst has a value on bar 27 and is 61936.8757259202 on bar 144008261a24)request.securityand is never calleddailyequals the control without the wrapper on all 1,441 bars08261a24)Case 3's numbers describe today's PineForge result only. The issue prints no TradingView value for it. Acceptance also does not mean there are no diagnostics: the scripts still report the applicable lookahead, EMA start-up and visual-only plot warnings.
This is fixed on main and ships in 1.11.0, whose publish target is 2026-11-10. This issue will be closed when that release is published.
Clarify unreleased fixes and numeric helper documentation (#353)
luisleo526 commentedon Nov 9, 2026
Fixed in v1.11.0: all four
request.securityforms in this issue are accepted and run in v1.11.0. Closing.