1
0
Fork 0
mirror of https://github.com/ruby/ruby.git synced 2022-11-09 12:17:21 -05:00
ruby--ruby/spec/ruby/core/string/index_spec.rb
eregon 401b64c4e8 Update to ruby/spec@c1b568b
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@62656 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2018-03-04 15:09:32 +00:00

316 lines
9.4 KiB
Ruby
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# -*- encoding: utf-8 -*-
require_relative '../../spec_helper'
require_relative 'fixtures/classes'
describe "String#index" do
it "raises a TypeError if passed nil" do
lambda { "abc".index nil }.should raise_error(TypeError)
end
it "raises a TypeError if passed a boolean" do
lambda { "abc".index true }.should raise_error(TypeError)
end
it "raises a TypeError if passed a Symbol" do
lambda { "abc".index :a }.should raise_error(TypeError)
end
it "calls #to_str to convert the first argument" do
char = mock("string index char")
char.should_receive(:to_str).and_return("b")
"abc".index(char).should == 1
end
it "calls #to_int to convert the second argument" do
offset = mock("string index offset")
offset.should_receive(:to_int).and_return(1)
"abc".index("c", offset).should == 2
end
it "raises a TypeError if passed a Fixnum" do
lambda { "abc".index 97 }.should raise_error(TypeError)
end
end
describe "String#index with String" do
it "behaves the same as String#index(char) for one-character strings" do
"blablabla hello cruel world...!".split("").uniq.each do |str|
chr = str[0]
str.index(str).should == str.index(chr)
0.upto(str.size + 1) do |start|
str.index(str, start).should == str.index(chr, start)
end
(-str.size - 1).upto(-1) do |start|
str.index(str, start).should == str.index(chr, start)
end
end
end
it "returns the index of the first occurrence of the given substring" do
"blablabla".index("").should == 0
"blablabla".index("b").should == 0
"blablabla".index("bla").should == 0
"blablabla".index("blabla").should == 0
"blablabla".index("blablabla").should == 0
"blablabla".index("l").should == 1
"blablabla".index("la").should == 1
"blablabla".index("labla").should == 1
"blablabla".index("lablabla").should == 1
"blablabla".index("a").should == 2
"blablabla".index("abla").should == 2
"blablabla".index("ablabla").should == 2
end
it "doesn't set $~" do
$~ = nil
'hello.'.index('ll')
$~.should == nil
end
it "ignores string subclasses" do
"blablabla".index(StringSpecs::MyString.new("bla")).should == 0
StringSpecs::MyString.new("blablabla").index("bla").should == 0
StringSpecs::MyString.new("blablabla").index(StringSpecs::MyString.new("bla")).should == 0
end
it "starts the search at the given offset" do
"blablabla".index("bl", 0).should == 0
"blablabla".index("bl", 1).should == 3
"blablabla".index("bl", 2).should == 3
"blablabla".index("bl", 3).should == 3
"blablabla".index("bla", 0).should == 0
"blablabla".index("bla", 1).should == 3
"blablabla".index("bla", 2).should == 3
"blablabla".index("bla", 3).should == 3
"blablabla".index("blab", 0).should == 0
"blablabla".index("blab", 1).should == 3
"blablabla".index("blab", 2).should == 3
"blablabla".index("blab", 3).should == 3
"blablabla".index("la", 1).should == 1
"blablabla".index("la", 2).should == 4
"blablabla".index("la", 3).should == 4
"blablabla".index("la", 4).should == 4
"blablabla".index("lab", 1).should == 1
"blablabla".index("lab", 2).should == 4
"blablabla".index("lab", 3).should == 4
"blablabla".index("lab", 4).should == 4
"blablabla".index("ab", 2).should == 2
"blablabla".index("ab", 3).should == 5
"blablabla".index("ab", 4).should == 5
"blablabla".index("ab", 5).should == 5
"blablabla".index("", 0).should == 0
"blablabla".index("", 1).should == 1
"blablabla".index("", 2).should == 2
"blablabla".index("", 7).should == 7
"blablabla".index("", 8).should == 8
"blablabla".index("", 9).should == 9
end
it "starts the search at offset + self.length if offset is negative" do
str = "blablabla"
["bl", "bla", "blab", "la", "lab", "ab", ""].each do |needle|
(-str.length .. -1).each do |offset|
str.index(needle, offset).should ==
str.index(needle, offset + str.length)
end
end
end
it "returns nil if the substring isn't found" do
"blablabla".index("B").should == nil
"blablabla".index("z").should == nil
"blablabla".index("BLA").should == nil
"blablabla".index("blablablabla").should == nil
"blablabla".index("", 10).should == nil
"hello".index("he", 1).should == nil
"hello".index("he", 2).should == nil
"Ive got a multibyte character.\n".index("\n\n").should == nil
end
with_feature :encoding do
it "returns the character index of a multibyte character" do
"ありがとう".index("").should == 2
end
it "returns the character index after offset" do
"われわれ".index("", 1).should == 2
end
it "returns the character index after a partial first match" do
"</</h".index("</h").should == 2
end
it "raises an Encoding::CompatibilityError if the encodings are incompatible" do
char = "".encode Encoding::EUC_JP
lambda do
"あれ".index char
end.should raise_error(Encoding::CompatibilityError)
end
end
end
describe "String#index with Regexp" do
it "behaves the same as String#index(string) for escaped string regexps" do
["blablabla", "hello cruel world...!"].each do |str|
["", "b", "bla", "lab", "o c", "d."].each do |needle|
regexp = Regexp.new(Regexp.escape(needle))
str.index(regexp).should == str.index(needle)
0.upto(str.size + 1) do |start|
str.index(regexp, start).should == str.index(needle, start)
end
(-str.size - 1).upto(-1) do |start|
str.index(regexp, start).should == str.index(needle, start)
end
end
end
end
it "returns the index of the first match of regexp" do
"blablabla".index(/bla/).should == 0
"blablabla".index(/BLA/i).should == 0
"blablabla".index(/.{0}/).should == 0
"blablabla".index(/.{6}/).should == 0
"blablabla".index(/.{9}/).should == 0
"blablabla".index(/.*/).should == 0
"blablabla".index(/.+/).should == 0
"blablabla".index(/lab|b/).should == 0
not_supported_on :opal do
"blablabla".index(/\A/).should == 0
"blablabla".index(/\Z/).should == 9
"blablabla".index(/\z/).should == 9
"blablabla\n".index(/\Z/).should == 9
"blablabla\n".index(/\z/).should == 10
end
"blablabla".index(/^/).should == 0
"\nblablabla".index(/^/).should == 0
"b\nablabla".index(/$/).should == 1
"bl\nablabla".index(/$/).should == 2
"blablabla".index(/.l./).should == 0
end
it "sets $~ to MatchData of match and nil when there's none" do
'hello.'.index(/.(.)/)
$~[0].should == 'he'
'hello.'.index(/not/)
$~.should == nil
end
it "starts the search at the given offset" do
"blablabla".index(/.{0}/, 5).should == 5
"blablabla".index(/.{1}/, 5).should == 5
"blablabla".index(/.{2}/, 5).should == 5
"blablabla".index(/.{3}/, 5).should == 5
"blablabla".index(/.{4}/, 5).should == 5
"blablabla".index(/.{0}/, 3).should == 3
"blablabla".index(/.{1}/, 3).should == 3
"blablabla".index(/.{2}/, 3).should == 3
"blablabla".index(/.{5}/, 3).should == 3
"blablabla".index(/.{6}/, 3).should == 3
"blablabla".index(/.l./, 0).should == 0
"blablabla".index(/.l./, 1).should == 3
"blablabla".index(/.l./, 2).should == 3
"blablabla".index(/.l./, 3).should == 3
"xblaxbla".index(/x./, 0).should == 0
"xblaxbla".index(/x./, 1).should == 4
"xblaxbla".index(/x./, 2).should == 4
not_supported_on :opal do
"blablabla\n".index(/\Z/, 9).should == 9
end
end
it "starts the search at offset + self.length if offset is negative" do
str = "blablabla"
["bl", "bla", "blab", "la", "lab", "ab", ""].each do |needle|
(-str.length .. -1).each do |offset|
str.index(needle, offset).should ==
str.index(needle, offset + str.length)
end
end
end
it "returns nil if the substring isn't found" do
"blablabla".index(/BLA/).should == nil
"blablabla".index(/.{10}/).should == nil
"blaxbla".index(/.x/, 3).should == nil
"blaxbla".index(/..x/, 2).should == nil
end
it "returns nil if the Regexp matches the empty string and the offset is out of range" do
"ruby".index(//,12).should be_nil
end
it "supports \\G which matches at the given start offset" do
"helloYOU.".index(/\GYOU/, 5).should == 5
"helloYOU.".index(/\GYOU/).should == nil
re = /\G.+YOU/
# The # marks where \G will match.
[
["#hi!YOUall.", 0],
["h#i!YOUall.", 1],
["hi#!YOUall.", 2],
["hi!#YOUall.", nil]
].each do |spec|
start = spec[0].index("#")
str = spec[0].delete("#")
str.index(re, start).should == spec[1]
end
end
it "converts start_offset to an integer via to_int" do
obj = mock('1')
obj.should_receive(:to_int).and_return(1)
"RWOARW".index(/R./, obj).should == 4
end
with_feature :encoding do
it "returns the character index of a multibyte character" do
"ありがとう".index(//).should == 2
end
it "returns the character index after offset" do
"われわれ".index(//, 1).should == 2
end
it "treats the offset as a character index" do
"われわわれ".index(//, 3).should == 3
end
it "raises an Encoding::CompatibilityError if the encodings are incompatible" do
re = Regexp.new "".encode(Encoding::EUC_JP)
lambda do
"あれ".index re
end.should raise_error(Encoding::CompatibilityError)
end
end
end