[WIP] catching Signal::INT #41
Conversation
|
Ok, update on this. I'm now setting a state on each command added to the command stack. The idea is when catching ^C, I can look at the command stack, and pop off the end all commands that are in a The issue I'm running in to now is that unexpected_eof errors don't push to the command stack immediately. So for these ones, I don't know their state. If I push these to the stack, then a lot of other weird things start to happen. I'll keep plugging away. |
…o parse control characters. Need to figure out the event loop issue
|
ICR stucks when I am pressing |
There was a problem hiding this comment.
Sorry for delay. I left there couple of comments.
It feels like to implement this feature we need to introduce new entity, something like ComposedCommand or something. But I am not sure that I want to dig to deep into it right now.
I hope sooner or later Crystal community will have a proper REPL without side effects)
src/icr/command_stack.cr
Outdated
| # returns `self`. Removes all values from the end of @commands where the state is `:err`. | ||
| def reset! | ||
| new_stack = commands.dup | ||
| commands.reverse.each { |c| c.state == "err" ? new_stack.pop : break } |
There was a problem hiding this comment.
What is the purpose of this method? I looks like your intention is to pop last command if it was erroneous (maybe I miss something), cause with first non-erroneous command break ends iteration.
If so, why not to use simple if condition?
There was a problem hiding this comment.
The idea behind this method is a "rollback". Each line you enter is like a commit. When you press ^C, it should rollback to the last good state. We need to keep the order in which things were entered, so I imagine the __exec_icr__ function would look something like...
1 + 1 # state is ok
a = 4 # state is ok
class Test # state is not ok
ebd # state is not ok
then ^C is pressed and it looks like this.
1 + 1 # state is ok
a = 4 # state is ok
src/icr/console.cr
Outdated
| @@ -51,7 +59,7 @@ module Icr | |||
| process_command(new_command) | |||
| when :error | |||
| # Give it the second try, validate the command in scope of entire file | |||
| @command_stack.push(command) | |||
| @command_stack.push(command + " ##{Icr::DELIMITER}err") | |||
There was a problem hiding this comment.
Patching command this way, seems a little bit strange to me. I am afraid it can cause some underdesired side effects. There must be only 1 delimiter and result output.
There was a problem hiding this comment.
Yeah, I thought about this too. I wanted to use a delimiter to make it easy just in case someone randomly enters a comment in to ICR. I figured using the built in one would be ok. Maybe we could have a secondary delimiter?
|
I get the same result. It's something to do with the Readline library that's causing that. If I switch to using a simple |
|
Just wanted to update on this (I haven't ditched it). I found a bug in crystal. Or, at least I'm thinking it's a bug.... I think there's a lot more Readline can do for us in this to make things a lot easier, but Crystal is missing a lot of what the readline lib has. |
|
I would prefer not to implement/merge this. ICR currently is full of tricks and glitches. And this PR only complicates things. |
|
Yeah, I agree. Crystal needs a bit more added to the Readline library to do this properly. I'll close this out for now, and we can touch back on it later. |
This PR isn't ready yet, but as an initial start I wanted to get your thoughts and input on some stuff.
Before this, ^C and ^D were caught by the same
process_inputwhich would just exit the console. Now, ^D will exit (just like IRB), but ^C is caught. The next step will be to create a state, and manage the state of the input. Then using ^C to reset the state to the last known clean state.Related: #33