Re: [RFC] Io\Terminal
| From: | Pratik Bhujel | Date: | Fri, 20 Nov 2026 09:51:59 +0000 |
| Subject: | Re: [RFC] Io\Terminal | ||
| References: | 1 2 3 4 | Groups: | php.internals |
| Request: | Send a blank email to internals+get-133136@lists.php.net to get a copy of this message | ||
Hi Tim,
Thanks for taking another look. I’ll hold off on the November 24 vote.
On reading the current raw-mode settings: you’re right, PHP can read
them again. I was wrong about that. The question for me is what a yes
or no would mean if echo is off but Ctrl+C still generates a signal.
There’s also a separate question for readLine(): does PHP hold a
lease? Reading the native settings doesn’t tell us that. I’d leave the
getter out for now and explain the distinction in the RFC.
On handling EOF separately: I’ll add a TerminalException subclass so
callers can handle the end of input without also catching an unrelated
I/O error. readLine() will keep returning null at EOF as it does now.
On the property name: agreed. I’ll use $columns for both the property
and the constructor parameter.
On future Key cases: your Ctrl+Left example makes sense. An
application could already be handling those bytes itself. Even if it
has a fallback for unknown enum cases, its existing Ctrl+Left handler
would stop matching. I’ll make clear in the RFC that changing a
sequence from a string to a Key case is a compatibility change.
I’ll post v0.7 here once the RFC and implementation are updated.
Regards,
Pratik