feat: Add functions for name mangling by TheZoq2 · Pull Request #170 · ethanuppal/marlin

TheZoq2 · 2025-12-12T10:41:07Z

No description provided.

ethanuppal

Thanks for this!

ethanuppal · 2025-12-12T10:54:58Z

+
+/// Performs the inverse of Verilator's name manglging
+/// https://verilator.org/guide/latest/languages.html#signal-naming
+pub fn demangle(s: &str) -> String {


for chars its fine but for strings I like to manually write out the full thing like string, which then usually means that my naming was bad and I should really name it something like name or identifier etc

ethanuppal · 2025-12-12T10:56:33Z

+
+/// Performs the inverse of Verilator's name manglging
+/// https://verilator.org/guide/latest/languages.html#signal-naming
+pub fn demangle(s: &str) -> String {


also function name could be better, maybe demangle_verilator_identifier or demangle_verilator_model_name or something idk?

ethanuppal · 2025-12-12T10:57:15Z

+                    } else {
+                        // We already checked that everything is ASCII, so this encoding trick
+                        // will not panic
+                        let mut buf = [0];


id write it as buffer

ethanuppal · 2025-12-12T10:57:53Z

+                result
+            })
+            .collect::<Vec<_>>()
+            .join("_05F"))


could you write this as computed from _'s ascii manually instead of leaving it magic?

if you don't want to do this, comment still not needed. You can make a constant called VERILATOR_MANGLED_UNDERSCORE: &'static str

ethanuppal · 2025-12-12T10:58:03Z

+            "{s} is a non-ascii name which is unsupported for name demangling."
+        ))
+    } else {
+        // Every character _except_ double _ can be handled as a single c haracter, so we'll


comment not really needed I think

I personally think it clarifies things

ethanuppal · 2025-12-12T10:58:58Z


+
+/// Performs Verilator's name manglging https://verilator.org/guide/latest/languages.html#signal-naming
+pub fn mangle(s: &str) -> Result<String, String> {


could you figure out how to make the error type more semantic than String? there are a bunch of ways and I trust your judgment

ethanuppal · 2025-12-12T11:01:57Z

+                    if c.is_ascii_alphanumeric() || c == '_' {
+                        result.push(c);
+                    } else {
+                        // We already checked that everything is ASCII, so this encoding trick


trying to figure out how to get rid of this comment; maybe you could just make the buffer have length 4, or have some documentating code like assert(c.is_ascii()? or maybe its fine to just remove it because people know the spec of encode_utf8

ethanuppal · 2025-12-12T11:02:33Z

+        // split on those, and then join them with their replacement
+        Ok(s.split("__")
+            .map(|segment| {
+                let mut result = String::new();


would it be uglier to write this as a fold instead of a stateful for loop?

I tend to prefer for loops over folds personally, it feels a bit more idiomatic to me somehow

do you think ::new or ::default reads better here

ethanuppal · 2026-02-27T10:36:06Z

    ffi_names::{DPI_INIT_CALLBACK, TRACE_EVER_ON},
 };

+#[derive(Debug, Snafu)]


Lol you're ahead of me, see #187, ideally I do this all in one PR and this can be an enum case

What do you want me to do with this one then? whatever!()?

Yeah for now unfortunately, it sucks but I think it's better to have consistency

ethanuppal · 2026-02-27T10:37:40Z

+                result
+            })
+            .collect::<Vec<_>>()
+            .join("_05F"))


if you don't want to do this, comment still not needed. You can make a constant called VERILATOR_MANGLED_UNDERSCORE: &'static str

ethanuppal · 2026-02-27T10:38:20Z

+    NonAsciiName { name: String },
+}
+
+/// Performs Verilator's name manglging https://verilator.org/guide/latest/languages.html#signal-naming


typos

all doc comments should end with periods

can you [name mangling](link)?

ethanuppal · 2026-02-27T10:38:38Z

+    }
+}
+
+/// Performs the inverse of Verilator's name manglging


typos

all doc comments should end with periods

can you [name mangling](link)

ethanuppal · 2026-02-27T10:39:20Z

+/// Performs the inverse of Verilator's name manglging
+/// https://verilator.org/guide/latest/languages.html#signal-naming
+pub fn demangle_verilator_name(name: &str) -> String {
+    name.split("__")


can this be a const as well? it's unclear why this is different from "_05F"

ethanuppal · 2026-02-27T10:39:47Z

+        .collect::<Vec<_>>()
+        .join("")
+}
+


Can you add idempotency tests and small unit tests?

ethanuppal · 2026-02-27T10:40:31Z

+                    if c.is_ascii_alphanumeric() || c == '_' {
+                        result.push(c);
+                    } else {
+                        // We already checked that everything is ASCII, so this


Comment not needed

Are you sure? It does hint that the following code can panic, but will not panic under current conditions

I think it's fine to assume anyone reading the code has perfect knowledge if the standard library docs, so you don't need to replicate that documentation

ethanuppal · 2026-02-27T10:41:37Z

+/// Performs Verilator's name manglging https://verilator.org/guide/latest/languages.html#signal-naming
+pub fn mangle_verilator_name(name: &str) -> Result<String, ManglingError> {
+    if name.is_ascii() {
+        // Every character _except_ double underscore can be handled as a single


Comment not needed, interested people can refer to the link you put

I don't think it is good to have to read another document for an explanation of why things are done if we can put the information right here.

This also prevents issues if upstream would move their docs around or something

ethanuppal · 2026-02-27T10:42:34Z

+/// https://verilator.org/guide/latest/languages.html#signal-naming
+pub fn demangle_verilator_name(name: &str) -> String {
+    name.split("__")
+        .map(|s| {


maybe segment instead of s?

I'm not sure, I think segment implies something more than it is (a split result where the start is an escapee, and the rest is normal text

TheZoq2 · 2026-02-27T12:29:15Z

Added some tests, ended up refactoring it a bit because I re-read the docs and saw some stuff I missed before

ethanuppal

Review

ethanuppal · 2026-02-27T13:43:00Z

+    }
+
+    #[test]
+    fn name_mangling_and_demangling_is_noop() {


id call it idempotent not noop
might be worth doing a simple fuzz as well with strings guaranteed to include double idents or symbols, but it is a simple enough function that it might not be needed. Your call

ethanuppal · 2026-02-27T13:48:00Z

+    if name.is_ascii() {
+        // Every character _except_ double underscore can be handled as a single
+        // character, so we'll split on those, and then join them with
+        // their replacement


Period at end of sentence

ethanuppal · 2026-02-27T13:48:22Z

+        // character, so we'll split on those, and then join them with
+        // their replacement
+        Ok(name
+            .split("__")


Also should be constant

ethanuppal · 2026-02-27T13:48:58Z

+pub fn demangle_verilator_name(name: &str) -> String {
+    name.split(VERILATOR_MANGLED_PREFIX)
+        .enumerate()
+        .map(|(i, s)| {


ethanuppal · 2026-02-27T13:54:34Z

        .args(["--lib-create", &library_name])
        .args(["--Mdir", verilator_artifact_directory.as_str()])
-        .args(["--top-module", top_module])
+        .args(["--top-module", &demangle_verilator_name(top_module)])


Wouldn't this be mangle not demangle? verilator/verilator#6940

github-actions Bot added the Code: Verilator label Dec 12, 2025

TheZoq2 changed the title ~~Add functions for name mangling and --no-trace-top~~ Add functions for name mangling Dec 12, 2025

TheZoq2 force-pushed the mangling branch 2 times, most recently from b8c9eb6 to 3f3adbb Compare December 12, 2025 10:46

TheZoq2 changed the title ~~Add functions for name mangling~~ [feat] Add functions for name mangling Dec 12, 2025

TheZoq2 changed the title ~~[feat] Add functions for name mangling~~ feat: Add functions for name mangling Dec 12, 2025

TheZoq2 force-pushed the mangling branch 2 times, most recently from 9c73c17 to d4503f7 Compare December 12, 2025 10:50

ethanuppal requested changes Dec 12, 2025

View reviewed changes

TheZoq2 force-pushed the mangling branch 3 times, most recently from f72dda9 to 4fc94c5 Compare February 27, 2026 10:24

ethanuppal requested changes Feb 27, 2026

View reviewed changes

TheZoq2 force-pushed the mangling branch from 4fc94c5 to a0a6c64 Compare February 27, 2026 12:09

Add Verilator name mangling

9bdec38

TheZoq2 force-pushed the mangling branch from a0a6c64 to 9bdec38 Compare February 27, 2026 12:25

Placeholder fix for Expression::Inside to get things to compile

a176148

github-actions Bot added the Code: Verilog Support label Apr 12, 2026

ethanuppal requested changes May 19, 2026

View reviewed changes

Fix review comments

8d57b9a



		/// Performs Verilator's name manglging https://verilator.org/guide/latest/languages.html#signal-naming
		pub fn mangle(s: &str) -> Result<String, String> {

Uh oh!

Conversation

TheZoq2 commented Dec 12, 2025

Uh oh!

ethanuppal left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

TheZoq2 commented Feb 27, 2026

Uh oh!

ethanuppal left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees