refactor: FFI #7
Conversation
Signed-off-by: lxl66566 <[email protected]>
Signed-off-by: lxl66566 <[email protected]>
| assert_eq!(gen.alloc_new_generator(rt.create_node().build()), 1); | ||
| } | ||
| } | ||
| mod tests {} |
There was a problem hiding this comment.
This pattern has been discarded. #9 shows the new generator id allocation scheme.
| @@ -164,6 +164,16 @@ mod tests { | |||
| use super::*; | |||
| use crate::{cljeval, init_jvm}; | |||
|
|
|||
| /// We can define a function in namespace, and call it later. | |||
| #[test] | |||
| fn test_defn_in_ns() -> Result<(), Box<dyn std::error::Error>> { | |||
There was a problem hiding this comment.
This test case is meaningless because you print the y, not asserting it.
There was a problem hiding this comment.
That's a test shows that two cljeval!() will share the same namespace.
If the hypothesis is not true, the jvm will panic so this test will fail.
Do you think the function name or the comment need to be changed?
Signed-off-by: lxl66566 <[email protected]>
Signed-off-by: lxl66566 <[email protected]>
Signed-off-by: lxl66566 <[email protected]>
Signed-off-by: lxl66566 <[email protected]>
Signed-off-by: lxl66566 <[email protected]>
| } | ||
|
|
||
| /// print a clojure instance | ||
| pub fn print_clj(inst: Instance) { |
There was a problem hiding this comment.
Why do we need a print_clj when we already have print?
There was a problem hiding this comment.
In some case, a clojure structure may not be print prettily by java.lang.System.out, it would be printed like CustomObject@6d06d69c.
But I'm sorry I cannot reproduce that currently
| with_jvm(|jvm| -> jResult<_> { jvm.to_rust(jvm.cast(inst, "java.lang.String")?) }) | ||
| } | ||
|
|
||
| /// This trait is for printing error messages better than `unwrap()` |
There was a problem hiding this comment.
except()/unwrap() will print message like that:
thread 'ffi::tests::mytest' panicked at src/ffi/mod.rs:248:49:
called `Result::unwrap()` on an `Err` value: JavaError("java.lang.IllegalStateException: Attempting to call unbound fn: #'jepsen.history/histo\n\tat clojure.lang.Var$Unbound.throwArity(Var.java:45)\n\tat clojure.lang.AFn.invoke(AFn.java:32)\n\tat clojure.lang.Var.invoke(Var.java:384)\n\tat java.base/jdk.internal.reflect.DirectMethodHandleAccessor.invoke(DirectMethodHandleAccessor.java:103)\n\tat java.base/java.lang.reflect.Method.invoke(Method.java:580)\n\tat org.astonbitecode.j4rs.api.invocation.JsonInvocationImpl.invokeMethod(JsonInvocationImpl.java:198)\n\tat org.astonbitecode.j4rs.api.invocation.JsonInvocationImpl.invoke(JsonInvocationImpl.java:70)\n")
The java error is wrapped as String by j4rs.
update: I modified the die function to print better string.
Signed-off-by: lxl66566 <[email protected]>
Signed-off-by: lxl66566 <[email protected]>
This PR is a recommitted PR, split from total modification.
utils.rstoffi/mod.rsNsRegisterto execute any clojure code from user.serde.clj, it provides customized ser/de methods. It's a fixup of serialization. (close Clojure ser/de inconsistency #10 ):for keytype objects. For example,{:a :b}will be translated to{a b}byorg.clojure/data.json.:key-fn keyword, the result will still be translated to{:a b}, the keytype is only for keys, not values.lib.rs. The previous tests inlib.rswere used to verify feasibility and are now temporarily deleted, after which they will be moved to the accordding modules.