string_value of DeriveActiveEnum errors on underscores, spaces and Non-Unicode UAX#31 chars

[denwong47]

Description

This Issue accompanies a PR that fixes it. PR Link to follow in comments below.

When using DeriveActiveEnum, struct ActiveEnum will try to reformat the string_value into an identifier for an Enum. Its does this by:

• if string_value starts with a number, prepend it with an underscore;
• otherwise, camel_case it.

This causes a few problems:

• the identifier will be an empty string if
  • string_value is an empty string, or
  • contains only non-XID_Start/XID_Continue characters, e.g. "$" or multiples of them
  • contains only characters that camel_case will remove: e.g. "_" or " "
• two string_value s that differ only by the above characters will become conflicting identifiers:
  • string_value of "fooBar" and "foo_bar" in the same enum will result conflicting identifiers of FooBar.
• if string_value leads with a digit, camel_case will NOT be applied, hence string_value = "0spaced word" will cause an error.

Steps to Reproduce

Scenario 1 - non-empty string_value results in empty identifier

use sea_orm::entity::prelude::*;

#[derive(Clone, Debug, PartialEq, EnumIter, DeriveActiveEnum)]
#[sea_orm(
    rs_type = "String",
    db_type = "String(None)",
    enum_name = "conflicting_string_values"
)]
pub enum ConflictingStringValues {
    #[sea_orm(string_value = "$")]
    Member1,
}

Scenario 2 - distinct string_value s result in conflict

use sea_orm::entity::prelude::*;

#[derive(Clone, Debug, PartialEq, EnumIter, DeriveActiveEnum)]
#[sea_orm(
    rs_type = "String",
    db_type = "String(None)",
    enum_name = "conflicting_string_values"
)]
pub enum ConflictingStringValues {
    #[sea_orm(string_value = "aB")]
    Member1,
    #[sea_orm(string_value = "a_b")]
    Member2,
}

Scenario 3 - string_value with leading numeric and includes spaces

use sea_orm::entity::prelude::*;

#[derive(Clone, Debug, PartialEq, EnumIter, DeriveActiveEnum)]
#[sea_orm(
    rs_type = "String",
    db_type = "String(None)",
    enum_name = "conflicting_string_values"
)]
pub enum ConflictingStringValues {
    #[sea_orm(string_value = "0 123")]
    Member1,
}

Expected Behavior

Compiles successfully with an Member1 matching string_value mapped.

Actual Behavior

Scenario 1

proc_macro panicked:
message: "" is not a valid identifier

Scenario 2

error[E0428]:
the name AB is defined multiple times

Scenario 3

proc_macro panicked:
message: "_0 123" is not a valid identifier

Reproduces How Often

Always reproducible.

Versions

sea-orm-macros 0.10.3

Additional Information

[denwong47]

PR open above.