search

rems-project / sail

Sail architecture definition language
Other
563 stars 92 forks source link

operators are not documented #605

Open ThinkOpenly opened 4 days ago

ThinkOpenly commented 4 days ago

Section "Operators" lists 15 built-in operators in a precedence table, but not their function. These should be documented, even if well known to practitioners in the art of programming.

In addition, the String type is given pretty short shrift:

CAUTION: Sail is not a language designed for working with strings, and the semantics of ISA specifications should not depend on any logic involving strings. If you find yourself using strings for reasons other than printing or logging errors in a Sail specification, you should probably reconsider.

(Should that text be maintained?)

However, there is at least one operator for strings which isn't even mentioned: ^ for concatenation. Or is that not in Sail proper? Or not yet? 0.18?

Alasdair commented 4 days ago

Strings don't work well in the SMT and SystemVerilog translations, which is why we want to discourage their use.

Ideally we would have a static analysis that checks that no logic involving strings ever influences control flow. Also making things 'stringly-typed' (as per this blog post: https://www.hanselman.com/blog/stringly-typed-vs-strongly-typed) when an enum/union would be a better choice is a common beginner mistake.

The string support in the library is super bare-bones, so I think the operator is defined by the RISC-V specification.