Thanks for fixing #386, however as it turns out, my attempt to simplify the demonstration code revealed unrelated problem, and my original problem was not fixed. It looks more related to ? now, but I'm no longer sure.
Here's more elaborate demonstration, closer to my code:
use askama::Template;
use serde_json::Value;
#[derive(Template)]
#[template(
ext = "html",
source = "
{%- macro p(v) %}{% if v == \"foo\" %}{% endif %}{% endmacro -%}
{%- if v.as_str().unwrap() == \"foo\" %}{% endif -%}
{%- call p(v.as_str().unwrap()) -%}
{%- if v.as_str().ok_or(\"error\")? == \"foo\" %}{% endif -%}
{%- call p(v.as_str().ok_or(\"error\")?) -%}
{%- let v1 = v.as_str().unwrap() %}
{%- if v1 == \"foo\" %}{% endif -%}
{%- call p(v1) -%}
{%- let v2 = v.as_str().ok_or(\"error\")? %}
{%- if v2 == \"foo\" %}{% endif -%}
{%- call p(v2) -%}
"
)]
struct Test {
v: serde_json::Value,
}
fn main() {
println!("{}", Test{v: Value::String("foo".to_string())}.render().unwrap());
}like the previous example, this worked with askama 0.12 - all cases were treated as &str.
With askama 0.13.0 (both release and master after #386 fix), however, commented out lines do not compile, because an extra reference is added:
use askama::Template;
use serde_json::Value;
#[derive(Template)]
#[template(
ext = "html",
source = "
{%- macro p(v) %}{% if v == \"foo\" %}{% endif %}{% endmacro -%}
{%- if v.as_str().unwrap() == \"foo\" %}{% endif -%}
{%- call p(v.as_str().unwrap()) -%}
{%- if v.as_str().ok_or(\"error\")? == \"foo\" %}{% endif -%}
{#%- call p(v.as_str().ok_or(\"error\")?) -%#}
{%- let v1 = v.as_str().unwrap() %}
{%- if v1 == \"foo\" %}{% endif -%}
{%- call p(v1) -%}
{%- let v2 = v.as_str().ok_or(\"error\")? %}
{#%- if v2 == \"foo\" %}{% endif -%#}
{#%- call p(v2) -%#}
"
)]
struct Test {
v: serde_json::Value,
}
fn main() {
println!("{}", Test{v: Value::String("foo".to_string())}.render().unwrap());
}So now it looks more like the problem is with ? operator, which, the same way as unwrap, returns &str, but it's mistreated and another reference is added both when it's passed to macro and when it's used in let expression (but not when it's used in expression directly).
Thanks for fixing #386, however as it turns out, my attempt to simplify the demonstration code revealed unrelated problem, and my original problem was not fixed. It looks more related to
?now, but I'm no longer sure.Here's more elaborate demonstration, closer to my code:
like the previous example, this worked with askama 0.12 - all cases were treated as
&str.With askama 0.13.0 (both release and master after #386 fix), however, commented out lines do not compile, because an extra reference is added:
So now it looks more like the problem is with
?operator, which, the same way as unwrap, returns&str, but it's mistreated and another reference is added both when it's passed to macro and when it's used inletexpression (but not when it's used in expression directly).