`querystring.parse` decodes incorrect in a specific case

[zkd8907]

• Version: 8.0.0/8.1.2
• Platform: Windows 10 x64/Linux x64

At Node 7.* or before, the result of require('querystring').parse('a=%20+&') is { a: ' ' }. However, at Node 8.0.0 or 8.1.2, the result of require('querystring').parse('a=%20+&') is { a: '%20 ' }. %20 doesn't decode when it is before +.

[targos]

That behavior changed in #11234

/cc @mscdex

[TimothyGu]

Aside from the issue, it is generally recommended that you use the new URLSearchParams class that parses query strings the exact same way browsers do, which is also faster than querystring module in almost all cases.

[jsilveira]

After 5 hours of struggling with a an intermittent issue in the encoding of some of the POSTs received from our website, we tracked down the issue to this change in handling trailing spaces. The "random" nature of the issue was caused by the fact that some people submitted the form input with a trailing space.

So, is this a bug or an expected "changed behaviour" ???

[TimothyGu]

There are a lot more problematic cases

// str    v8.x     v6.x         expected
'af+'     {}       {}            { 'af ': '' }
'af+&'    {}       { 'af ': '' } { 'af ': '' }     REGRESSION
'%20+'    {}       {}            { '  ': '' }
'%20+&'   {}       { '  ': '' }  { '  ': '' }      REGRESSION
'+'       {}       {}            { ' ': '' }
'+&'      {}       { ' ': '' }   { ' ': '' }       REGRESSION

[TimothyGu]

@jsilveira It is a bug.