Upgrading Ruby 1.8.6 to 1.8.7

Although this post is a little over due I thought it could potentially help other people attempting to do the update. I wanted a guide which could help people do the update and understand what needs to go into doing it.

OSX Install

For some reason tcl-tk is borked with the Ruby 1.8.7. You can install 1.8.7 by doing the following.

rvm install 1.8.7 --without-tcl --without-tk
brew install tcl-tk
rvm reinstall 1.8.7

Searching Your Code Base

Ruby 1.8.7 mainly added additional methods to classes. This means that it is pretty much backwards compatible. There are a few exceptions regarding incompatibility. These exceptions are documented here.

String#slice!

String#slice! had some unintentional bugs and they have been fixed because either they disagreed with documentation or their respective behavior of #slice. Unfortunately, this causes some incompatibilities in the following (somewhat rare) cases.

• `#slice!` no longer expands the array when an out-of-boundary value is given.

Ruby 1.8.6

a = [1,2]
a.slice!(4,0)   #=> nil
a               #=> [1,2,nil,nil]

Ruby 1.8.7

a = [1,2]
a.slice!(4,0)   #=> nil
a               #=> [1,2]

• `#slice!` no longer raises an exception but returns nil when a negative length or out-of-boundary negative position is given.

Ruby 1.8.6

a = [1,2]
a.slice!(1,-1)  #=> (raises IndexError)
a.slice!(-5,1)  #=> (raises IndexError)

Ruby 1.8.7

a = [1,2]
a.slice!(1,-1)  #=> nil
a.slice!(-5,1)  #=> nil

String#to_i

The behavior in Ruby 1.8.6 was inconsistent with Ruby syntax and considered as a bug. String#to_i, String#hex and String#oct no longer accept a sequence of underscores (`__’) as part of a number.

Ruby 1.8.6

'1__0'.to_i     #=> 10
'1__0'.to_i(2)  #=> 2  # 0b10
'1__0'.oct      #=> 8  # 010
'1__0'.hex      #=> 16 # 0x10

Ruby 1.8.7

'1__0'.to_i     #=> 1
'1__0'.to_i(2)  #=> 1
'1__0'.oct      #=> 1
'1__0'.hex      #=> 1

Date.parse

’##.##.##’ (where each ‘#’ is a digit) is now taken as ‘YY.MM.DD’ instead of ‘MM.DD.YY’. While the change may confuse you, you can use .strptime to replicate the old way it was done.

Date.strptime("01.02.03", "%m.%d.%Y")

I could not actually replicate this functionality personally. Let me know if you have any success.

1.8.7 :001 > a = Date.parse("01.02.03")
=> #<Date: 3442913/2,0,2299161>

1.8.6 :001 > a = Date.parse("01.02.03")
=> #<Date: 3444309/2,0,2299161>

Keep in mind, this changed again in 1.9.

StringIO#each_byte

The return value changed from nil to self. This is what the document says and the same as each_line() does.

tempfile

The file name format has changed. No dots are included by default in temporary file names any more. This shouldn’t be a problem, unless you are permanently keeping temporary files. Which is the exact opposite of what temp files should be used for. Seems like a bad coding practice to me.

URI

Made changes to FTP and URI.merge to accommodate RFC3986. These are more bugs in 1.8.6 than anything else. But still search your code base to see if URI is used.

FTP

URI('ftp://example.com/foo').path #=> 'foo'
URI('ftp://example.com/%2Ffoo').path #=> '/foo'
URI::FTP.build([nil, 'example.com', nil, '/foo', 'i').to_s #=> 'ftp://example.com/%2Ffoo;type=i'

URI.merge

URI('http://a/b/c/d;p?q').merge('?y') == URI('http://a/b/c/d;p?y')
URI('http://a/b/c/d;p?q').merge('/./g') == URI('http://a/g')
URI('http://a/b/c/d;p?q').merge('/../g') == URI('http://a/g')
URI('http://a/b/c/d;p?q').merge('../../../g') == URI('http://a/g')
URI('http://a/b/c/d;p?q').merge('../../../../g') == URI('http://a/g')