AST transformer rewrites barstate.isconfirmed as a call, but historical-mode subprocess assigns it as a literal True → TypeError: 'bool' object is not callable

[Zombajo]

Summary

Filing as a follow-up to #49 (which you closed as not-a-bug because the request.security subprocess is historical-only by design). Agreed on that design intent — but the AST transformer rewrites every barstate.isconfirmed reference into a property call, and the historical-mode subprocess assigns the literal True to that name. The two are incompatible: any compiled strategy that references barstate.isconfirmed inside a code path reachable from the subprocess crashes with:

TypeError: 'bool' object is not callable

This is internal PyneCore behavior (AST transformer ↔ historical-mode subprocess) rather than a TV-divergence issue, which is why it deserves its own report.

Reproduction

Any compiled strategy that uses barstate.isconfirmed and also calls request.security will crash on the first subprocess iteration. Minimal Pine source:

//@version=6
strategy("isconfirmed_subprocess_crash")
htf_close = request.security(syminfo.tickerid, "60", close)
if barstate.isconfirmed
    strategy.entry("L", strategy.long)

Compile with PyneSys, run on any data. Expected: strategy evaluates barstate.isconfirmed correctly inside the subprocess. Observed: TypeError: 'bool' object is not callable on the first subprocess bar.

Suspected cause

The AST transformer (likely in the compile pipeline that PyneSys produces) rewrites barstate.isconfirmed into a property / method call site. The historical-mode subprocess module (around core/security_process.py:150-160 in our v6.4.5 install) assigns the literal True to that name in the subprocess namespace. Literal True isn't callable, so the rewritten call crashes.

Proposed fix direction

Either:

• The historical-mode subprocess assigns a callable wrapper (e.g., a lambda: True or a property descriptor) instead of the literal, OR
• The AST transformer detects historical-mode subprocess context and doesn't rewrite barstate.isconfirmed into a call there.

We've been running with a local patch at security_process.py:150-160 that swaps the literal for a callable _isconfirmed_true marker. With the patch removed (per your #49 verdict), our test pipeline crashes immediately on the first subprocess bar regardless of strategy logic.

Environment

• PyneCore version: pynesys-pynecore 6.4.5
• Python: 3.13.7
• OS: Windows 11

Related

• Percent commission underreported on entry fills — sizing path uses correct formula, execution path uses buggy formula #48 (percent-commission, fixed in v6.4.5 — thanks)
• barstate.isconfirmed returns wrong value inside request.security subprocess #49 (closed as not-a-bug — agree on design intent; this is the orthogonal AST/subprocess incompatibility)
• Strategy PnL output for futures contracts is in points, not USD — pointvalue not applied #50 (futures pointvalue, open)
• strategy.commission.cash_per_contract charges only entry, not exit #51 (cash_per_contract single-leg, open)

[wallneradam]

Thanks for the careful write-up and the standalone report — splitting it from #49 was the right call, the AST-vs-subprocess incompatibility really is an orthogonal bug.

This is already fixed on main (commit 9450260, "fix(barstate): expose runtime states as mutable variables"). The fix converts all seven lib.barstate.* entries from @module_property to plain bool variables and updates transformers/module_properties.json to mark them as "type": "variable", so the AST transformer no longer wraps barstate.isconfirmed (or any sibling state) in a () call. The literal-True assignment in security_process.py is now bool-to-bool, fully compatible.

The fix landed on main shortly before you filed this — it was originally done on another branch a while ago, but we ported it over today because that branch is still some way from being merged.

It will ship in v6.4.6. Closing as fixed; please reopen if anything related still misbehaves after the upgrade.