Compile-Time Detection

The @with_pool macro statically analyzes your code at macro expansion time — zero runtime cost, always active.

Escape Analysis (PoolEscapeError)

Detects any tail expression that would deliver a pool-backed array out of the scope. Tracks aliases, containers, and convenience wrappers. Severity is form-based, because the two forms differ in what the macro can know:

• Function form and explicit return (either form): the value definitively reaches the enclosing function's caller — a begin block is not a function boundary, so return inside a block form returns from the enclosing function. These always throw PoolEscapeError at expansion.
• Block form, implicit last expression: the macro cannot see its own call site, so whether the block's value is used (out = @with_pool ...) or discarded (a loop body, a bare statement) is undecidable. These emit a @warn and the expansion replaces the escaping value with an inert EscapedPoolArray guard: discarded → completely silent; used → the first operation throws EscapedPoolUseError with the variable name, shape, and scope location.

@with_pool pool function bad()
    A = acquire!(pool, Float64, 3, 3)
    return A         # ← PoolEscapeError: function return, always an error
end

out = @with_pool pool begin
    v = acquire!(pool, Float64, 100)
    w = v            # alias tracked
    w                # ← @warn + guard: `out` is an EscapedPoolArray
end
out[1]               # ← EscapedPoolUseError: names `w`, its shape, and the scope

# The guard is inert when the value is never used — this common
# copy-out-and-discard shape runs silently and correctly:
Threads.@threads for k in 1:nmodels
    @with_pool pool begin
        y = acquire!(pool, Float64, m, n)
        compute!(y, k)
        results[:, :, k] = y   # copy runs; the yielded `y` is guarded; @threads discards it
    end
end

Error messages include the variable name, declaration site, and escaping return expression:

ERROR: LoadError: PoolEscapeError (compile-time)

  The following variable escapes the @with_pool scope:

    v  ← pool-acquired array

  Declarations:
    [1]  v = acquire!(pool, Float64, 100)  [myfile.jl:2]

  Escaping return:
    [1]  v  [myfile.jl:3]

  Fix: Use collect(v) to return owned copies.
       Or use a regular Julia array (zeros()/Array{T}()) if it must outlive the pool scope.

in expression starting at myfile.jl:1

Incidental Tail Patterns

Beyond the direct-return patterns above, three additional tail shapes are checked — they don't look like a return, but the scope's last expression still evaluates to a pool-backed array, which then escapes as the scope's value:

@with_pool pool begin
    v = acquire!(pool, Float64, 100)
    acquire!(pool, Float64, 100)   # ← direct acquire-call tail
end

@with_pool pool begin
    v = acquire!(pool, Float64, 100)
    v .= 0.0                        # ← broadcast-assign tail (evaluates to `v`)
end

@with_pool pool begin
    v = acquire!(pool, Float64, 100)
    y = v                           # ← assignment tail
end

Fix: end the block with nothing if the value is meant to be discarded — this also silences the warning:

@with_pool pool begin
    v = acquire!(pool, Float64, 100)
    v .= 0.0
    nothing   # ← no longer escapes
end

In block form these follow the same warn + EscapedPoolArray-guard treatment as all implicit tails; in function form or under an explicit return, they throw. The escape_lint preference tunes the block-form severity:

• "warn" (default) — @warn diagnostic + the guard rewrite described above.
• "error" — block-form implicit tails also throw PoolEscapeError (the strict pre-guard behavior; no guard rewrite needed).
• "off" — no warning and no guard rewrite (the block tail escapes the raw pool array, as before this feature). Function-form and explicit-return direct escapes still always error.

using Preferences
Preferences.set_preferences!("AdaptiveArrayPools", "escape_lint" => "error")

Mutation Analysis (PoolMutationError)

Structural mutations (resize!, push!, append!, etc.) on pool-backed arrays are rejected:

@with_pool pool begin
    v = acquire!(pool, Float64, 100)
    push!(v, 1.0)   # ← PoolMutationError
end

Detected functions: resize!, push!, pop!, pushfirst!, popfirst!, append!, prepend!, insert!, deleteat!, splice!, sizehint!, empty!.

Limitations

Static analysis can't see through opaque function calls:

function sneaky(v)
    return v  # opaque — macro can't see inside
end

@with_pool pool begin
    v = acquire!(pool, Float64, 100)
    sneaky(v)  # escapes through opaque call — not caught at compile time
end

That's what the runtime layer is for.