Don't strip spaces for regular REPL input

[edef1c]

This means .method() isn't misparsed as a REPL keyword, and significantly improves the usability of .load.

[edef1c]

Expected (and this patch's) behaviour:

> .load test.js
> function sum(arr) {
...   return arr
...     .reduce(function(acc, x) { return acc + x }, 0)
... }
undefined
> console.log(sum([20, 1, 21]))
42
undefined
> 

Current behaviour:

> .load test.js
> function sum(arr) {
...   return arr
...     .reduce(function(acc, x) { return acc + x }, 0)
Invalid REPL keyword
undefined
> }
SyntaxError: Unexpected token }
    at Object.exports.createScript (vm.js:24:10)
    at REPLServer.defaultEval (repl.js:131:25)
    at bound (domain.js:254:14)
    at REPLServer.runBound [as eval] (domain.js:267:12)
    at REPLServer.<anonymous> (repl.js:308:12)
    at emitOne (events.js:82:20)
    at REPLServer.emit (events.js:169:7)
    at REPLServer.Interface._onLine (readline.js:209:10)
    at REPLServer.Interface._line (readline.js:548:8)
    at REPLServer.Interface._ttyWrite (readline.js:879:20)
> console.log(sum([20, 1, 21]))
ReferenceError: sum is not defined
    at repl:1:13
    at REPLServer.defaultEval (repl.js:154:27)
    at bound (domain.js:254:14)
    at REPLServer.runBound [as eval] (domain.js:267:12)
    at REPLServer.<anonymous> (repl.js:308:12)
    at emitOne (events.js:82:20)
    at REPLServer.emit (events.js:169:7)
    at REPLServer.Interface._onLine (readline.js:209:10)
    at REPLServer.Interface._line (readline.js:548:8)
    at REPLServer.Interface._ttyWrite (readline.js:879:20)
> 

[silverwind]

cc @thefourtheye

[Fishrock123]

Oops, probably my fault. Hmm, will need to fix this before the 2.5.0 release.

[edef1c]

@Fishrock123 Nah, this has been the case for years.

[thefourtheye]

Hmmm. I knew about this weak REPL command check when I submitted REPL bug fixes PR, but I could not come up with a practical usecase. So, apart from removing the check mentioned in this PR, it would be better if we improved the REPL command check with a regular expression. Instead of

cmd && cmd.charAt(0) === '.' && isNaN(parseFloat(cmd))

if we had

cmd && isNaN(parseFloat(cmd)) && /.[a-z]+\s*[a-z]*/.test(cmd)

it would be better. Also, I would recommend including a test as well.

Edit: To me, stripping spaces is not the actual problem here (even though leaving the spaces when we are not inside a string literal is not a big deal), but considering the current line of input as a REPL command just because it starts with . doesn't seem right. That is why I proposed the RegEx above.

Edit 2: It would be better if the RegEx is /^.[a-z]+\s*[\w]*$/.test(cmd)

[targos]

@thefourtheye you can easily run into this issue by copy-pasting something in the REPL

[thefourtheye]

@targos Yup, I understand that after seeing the example shown in the PR. That is why using a RegEx to validate the string should work here. What do you think?

[targos]

@thefourtheye Yes a RegExp could be the solution. We just need to make sure that it does not identify too much valid JS as a REPL command.

[jasnell]

@nathan7 @thefourtheye @Fishrock123 ... what's the status on this one?

[princejwesley]

@Fishrock123 @thefourtheye /^\.[a-z][^(]*$/ is sufficient, right?

[thefourtheye]

@princejwesley I proposed almost the same in #2254. But the problem is, load and save accept filenames as inputs following them. If they had ( in them then this solution will break.

[princejwesley]

@jasnell @thefourtheye #3835 fixed this issue.