Show short dump on commands which do not look like they were meant for inspection

[tgr]

Object dumps are great for inspecting state but when I am just setting up the state and performing a series of commands, each of which returns a complex object with a multi-page dump, that gets annoying fast. It would be nice if the interpreter would recognize which commands are not meant for inspection (commands including an assignment operator would be a decent first approximation) and use a short version of the dump (probably just the class name would work well enough) in that case.

#218 is sort of related.

[phil-s]

You can follow the command with another command. PsySH only prints the value of the last statement it executed. e.g.:

>>> thing_which_returns_a_lot_of_state();1;
1

The interpreter would be making quite the assumption to decide that a given result wasn't intended for inspection. There would need to be a way of ensuring that the full dump was at least optionally printed.

Is there currently a way to re-print the most recent return value? An interesting approach might be to print the "short version" by default, but have a command to re-print the last value in full. dump seems to require an argument, so re-printing the most recent value could become the behaviour when no argument was provided?

[tgr]

Is there currently a way to re-print the most recent return value?

Yeah, dump $_. But the more intuitive (to me) version is

$foo = someCommand();
$foo

where the second command can't have any other possible reason than inspection. Similarly a command like $foo['bar'][$baz] is no doubt used for inspection. So either using a full dump when there is no assignment, or using it when there is no assignment and no function/method call (ie. no side effect) feels natural to me.

You can follow the command with another command.

Thanks, that's a nice workaround.

[bobthecow]

The implicit dump for return values is supposed to be printing a short dump. The dump command is there if you want more info. If it's not doing that now, we should fix it :)