-
Notifications
You must be signed in to change notification settings - Fork 448
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add support for compound assignment operators #3669
Conversation
This patch adds support for compound assignment expressions of the form `E1 op= E2`. Signed-off-by: Radostin Stoyanov <radostin.stoyanov@eng.ox.ac.uk>
@@ -1161,6 +1173,54 @@ assignmentOrMethodCallStatement | |||
$1, $3, $6); | |||
$$ = new IR::MethodCallStatement(@1 + @7, mc); } | |||
|
|||
| lvalue "+=" expression ";" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I am not sure this is the right semantics. This implementation will repeat the side effects on the lhs twice.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Note most lvalue by definition could not have a side effect.
The only one which I can think of is: headerstack[f(a)].b
where f is defined to take 1 arguments which is inout (you could do something similar with externs too).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
regardless of the inout arguments, a method call may have arbitrary side effects within in, and the way this code duplicates the reference to the lhs, a method call inside an index would end up being called twice. We have a pass (DoSimplifyExpressions) that exists to deal with this kind of program, but there are a number of other passes that currently must run before it (such as typeChecking) that would need to understand what is going on. So this almost certainly needs a new IR node type (some kind of op-assignment statement) to represent this until SimplifyExpressions can turn it into a normal AssignmentStatement, and all the passes that run before need to know about it.
We will need to be careful if we want to ensure that l-values never have side effects. I am not even sure that's true of the current language design and it would be fragile to ensure it going forward. |
@rst0git are you still interested in completing this feature? Do you need help with the |
@rst0git, can you let us know if you are still working on this? We'd be glad to see it added, but need to fix the side-effect issue... |
@jnfoster I'm sorry for the delayed response. I will update the pull request in the next few days. |
Superceeded by #5108 |
This pull request adds support for compound assignment expressions of the form
E1 op= E2
as discussed in p4lang/p4-spec#1144 and p4lang/p4-spec#1139.Fixes #3258