Repository navigation
assigning _ in the repl has side effects #5431
Description
- Version: v4.2.3
- Platform: Linux eric-ThinkPad-X220 3.19.0-51-generic Fix for path.resolve containing relative path to another drive in windows #57~14.04.1-Ubuntu SMP Fri Feb 19 14:36:55 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux
In Nodejs REPL env, where var _ = require('lodash') failed to be assigned, but abc can.
> var _ = require('lodash');
undefined
> _
undefined
> var abc = require('lodash');
undefined
> abc
{ [Function: lodash]
templateSettings:
{ escape: /<%-([\s\S]+?)%>/g,
evaluate: /<%([\s\S]+?)%>/g,
interpolate: /<%=([\s\S]+?)%>/g,
variable: '',
imports: { _: [Circular] } },
Activity
mscdex commentedon Feb 25, 2016
In the
repldocumentation it mentions that the_variable is special and always refers to the result of the last expression. In other words, you cannot use_in the repl for your own use, as it will get overwritten every time you execute a "line" of code.calidion commentedon Feb 25, 2016
the use of
_is so common for lodash or underscore. I don't have any reason why_is chosen to be the default result..would be far more acceptable than_vkurchatkin commentedon Feb 25, 2016
. is not a valid identifier
rvagg commentedon Feb 25, 2016
Just out of interest I thought I've do some provenance work on this
_was introduced in repl.js back in the 25th of September 2009 @ 7674bd5Underscore was born almost exactly one month later, on the 26th of October 2009 @ jashkenas/underscore@02ede85
Perhaps we should go argue to Underscore and lodash that they should change to something that won't conflict with Node?
vkurchatkin commentedon Feb 25, 2016
FWIW Chrome console uses
$_for this and it works the same waybnoordhuis commentedon Feb 25, 2016
To add to what @rvagg said, the underscore's behavior was copied from the python and ruby REPLs.
calidion commentedon Feb 25, 2016
I think
__or__lastor__xxxwould be much acceptable where__is known to be system reserved.bnoordhuis commentedon Feb 25, 2016
It's been
_for over half a decade now, it's not going to change.vkurchatkin commentedon Feb 25, 2016
We could try two things:
_;_is defined;calidion commentedon Feb 25, 2016
@vkurchatkin
I would say that is a very good idea.
💯
rvagg commentedon Feb 25, 2016
I like that suggestion @vkurchatkin, would lead to fewer surprises like the one that prompted this issue
silverwind commentedon Feb 25, 2016
Why not just disable
_functionality with a warning once it's being assigned to? Not being able to paste lodash examples in the REPL is not ideal, imho.[-]strange bug[/-][+]assigning _ in the repl has side effects[/+]princejwesley commentedon Feb 25, 2016
I suggest to add an entry in
helpcommand like belowrvagg commentedon Feb 26, 2016
see #5438 for one option
repl: Assignment of _ allowed with warning