discourse/spec/components/text_cleaner_spec.rb

233 lines
8.0 KiB
Ruby
Raw Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

require 'rails_helper'
require 'text_cleaner'
describe TextCleaner do
context "exclamation marks" do
let(:duplicated_string) { "my precious!!!!" }
let(:deduplicated_string) { "my precious!" }
it "ignores multiple ! by default" do
expect(TextCleaner.clean(duplicated_string)).to eq(duplicated_string)
end
it "deduplicates ! when enabled" do
expect(TextCleaner.clean(duplicated_string, deduplicate_exclamation_marks: true)).to eq(deduplicated_string)
end
end
context "question marks" do
let(:duplicated_string) { "please help me????" }
let(:deduplicated_string) { "please help me?" }
it "ignores multiple ? by default" do
expect(TextCleaner.clean(duplicated_string)).to eq(duplicated_string)
end
it "deduplicates ? when enabled" do
expect(TextCleaner.clean(duplicated_string, deduplicate_question_marks: true)).to eq(deduplicated_string)
end
end
context "all upper case text" do
let(:all_caps) { "ENTIRE TEXT IS ALL CAPS" }
let(:almost_all_caps) { "ENTIRE TEXT iS ALL CAPS" }
let(:regular_case) { "entire text is all caps" }
it "ignores all upper case text by default" do
expect(TextCleaner.clean(all_caps)).to eq(all_caps)
end
it "replaces all upper case text with regular case letters when enabled" do
expect(TextCleaner.clean(all_caps, replace_all_upper_case: true)).to eq(regular_case)
end
it "ignores almost all upper case text when enabled" do
expect(TextCleaner.clean(almost_all_caps, replace_all_upper_case: true)).to eq(almost_all_caps)
end
end
context "first letter" do
let(:lowercased) { "this is awesome" }
let(:capitalized) { "This is awesome" }
let(:iletter) { "iLetter" }
it "ignores first letter case by default" do
expect(TextCleaner.clean(lowercased)).to eq(lowercased)
expect(TextCleaner.clean(capitalized)).to eq(capitalized)
expect(TextCleaner.clean(iletter)).to eq(iletter)
end
it "capitalizes first letter when enabled" do
expect(TextCleaner.clean(lowercased, capitalize_first_letter: true)).to eq(capitalized)
expect(TextCleaner.clean(capitalized, capitalize_first_letter: true)).to eq(capitalized)
expect(TextCleaner.clean(iletter, capitalize_first_letter: true)).to eq(iletter)
end
end
context "periods at the end" do
let(:with_one_period) { "oops." }
let(:with_several_periods) { "oops..." }
let(:without_period) { "oops" }
it "ignores unnecessary periods at the end by default" do
expect(TextCleaner.clean(with_one_period)).to eq(with_one_period)
expect(TextCleaner.clean(with_several_periods)).to eq(with_several_periods)
end
it "removes unnecessary periods at the end when enabled" do
expect(TextCleaner.clean(with_one_period, remove_all_periods_from_the_end: true)).to eq(without_period)
expect(TextCleaner.clean(with_several_periods, remove_all_periods_from_the_end: true)).to eq(without_period)
end
it "keeps trailing whitespaces when enabled" do
expect(TextCleaner.clean(with_several_periods + " ", remove_all_periods_from_the_end: true)).to eq(without_period + " ")
end
end
context "extraneous space" do
let(:with_space_exclamation) { "oops !" }
let(:without_space_exclamation) { "oops!" }
let(:with_space_question) { "oops ?" }
let(:without_space_question) { "oops?" }
it "ignores extraneous space before the end punctuation by default" do
expect(TextCleaner.clean(with_space_exclamation)).to eq(with_space_exclamation)
expect(TextCleaner.clean(with_space_question)).to eq(with_space_question)
end
it "removes extraneous space before the end punctuation when enabled" do
expect(TextCleaner.clean(with_space_exclamation, remove_extraneous_space: true)).to eq(without_space_exclamation)
expect(TextCleaner.clean(with_space_question, remove_extraneous_space: true)).to eq(without_space_question)
end
it "keep trailing whitespaces when enabled" do
expect(TextCleaner.clean(with_space_exclamation + " ", remove_extraneous_space: true)).to eq(without_space_exclamation + " ")
expect(TextCleaner.clean(with_space_question + " ", remove_extraneous_space: true)).to eq(without_space_question + " ")
end
end
context "interior spaces" do
let(:spacey_string) { "hello there's weird spaces here." }
let(:unspacey_string) { "hello there's weird spaces here." }
it "ignores interior spaces by default" do
expect(TextCleaner.clean(spacey_string)).to eq(spacey_string)
end
it "fixes interior spaces when enabled" do
expect(TextCleaner.clean(spacey_string, fixes_interior_spaces: true)).to eq(unspacey_string)
end
end
context "leading and trailing whitespaces" do
let(:spacey_string) { " \t test \n " }
let(:unspacey_string) { "test" }
it "ignores leading and trailing whitespaces by default" do
expect(TextCleaner.clean(spacey_string)).to eq(spacey_string)
end
it "strips leading and trailing whitespaces when enabled" do
expect(TextCleaner.clean(spacey_string, strip_whitespaces: true)).to eq(unspacey_string)
end
end
context "title" do
it "fixes interior spaces" do
expect(TextCleaner.clean_title("Hello there")).to eq("Hello there")
end
it "strips leading and trailing whitespaces" do
expect(TextCleaner.clean_title(" \t Hello there \n ")).to eq("Hello there")
end
it "strips zero width spaces" do
expect(TextCleaner.clean_title("Hello there")).to eq("Hello there")
expect(TextCleaner.clean_title("Hello there").length).to eq(11)
end
context "title_prettify site setting is enabled" do
before { SiteSetting.title_prettify = true }
it "deduplicates !" do
expect(TextCleaner.clean_title("Hello there!!!!")).to eq("Hello there!")
end
it "deduplicates ?" do
expect(TextCleaner.clean_title("Hello there????")).to eq("Hello there?")
end
it "replaces all upper case text with regular case letters" do
expect(TextCleaner.clean_title("HELLO THERE")).to eq("Hello there")
end
it "doesn't replace all upper case text when uppercase posts are allowed" do
SiteSetting.allow_uppercase_posts = true
expect(TextCleaner.clean_title("HELLO THERE")).to eq("HELLO THERE")
end
it "capitalizes first letter" do
expect(TextCleaner.clean_title("hello there")).to eq("Hello there")
end
it "removes unnecessary period at the end" do
expect(TextCleaner.clean_title("Hello there.")).to eq("Hello there")
end
it "removes extraneous space before the end punctuation" do
expect(TextCleaner.clean_title("Hello there ?")).to eq("Hello there?")
end
it "replaces all upper case unicode text with regular unicode case letters" do
expect(TextCleaner.clean_title("INVESTIGAÇÃO POLÍTICA NA CÂMARA")).to eq("Investigação política na câmara")
end
it "doesn't downcase text if only one word is upcase in a non-ascii alphabet" do
expect(TextCleaner.clean_title("«Эта неделя в EVE»")).to eq("«Эта неделя в EVE»")
end
it "capitalizes first unicode letter" do
expect(TextCleaner.clean_title("épico encontro")).to eq("Épico encontro")
end
end
end
describe "#normalize_whitespaces" do
it "normalize whitespaces" do
whitespaces = "\u0020\u00A0\u1680\u180E\u2000\u2001\u2002\u2003\u2004\u2005\u2006\u2007\u2008\u2009\u200A\u2028\u2029\u202F\u205F\u3000"
expect(whitespaces.strip).not_to eq("")
expect(TextCleaner.normalize_whitespaces(whitespaces).strip).to eq("")
expect(TextCleaner.normalize_whitespaces(nil)).to be_nil
end
it "does not muck with zero width white space" do
# this is used for khmer, dont mess with it
expect(TextCleaner.normalize_whitespaces("hello\u200Bworld").strip).to eq("hello\u200Bworld")
expect(TextCleaner.normalize_whitespaces("hello\uFEFFworld").strip).to eq("hello\uFEFFworld")
end
end
end