class YARP::LexCompat::Heredoc::DedentingHeredoc
Heredocs that are dedenting heredocs are a little more complicated. Ripper outputs on_ignored_sp tokens for the whitespace that is being removed from the output. YARP
only modifies the node itself and keeps the token the same. This simplifies YARP
, but makes comparing against Ripper much harder because there is a length mismatch.
Fortunately, we already have to pull out the heredoc tokens in order to insert them into the stream in the correct order. As such, we can do some extra manipulation on the tokens to make them match Ripper’s output by mirroring the dedent logic that Ripper uses.
Constants
- TAB_WIDTH
Attributes
Public Class Methods
# File yarp/lex_compat.rb, line 354 def initialize @tokens = [] @dedent_next = true @dedent = nil @embexpr_balance = 0 end
Public Instance Methods
As tokens are coming in, we track the minimum amount of common leading whitespace on plain string content tokens. This allows us to later remove that amount of whitespace from the beginning of each line.
# File yarp/lex_compat.rb, line 364 def <<(token) case token.event when :on_embexpr_beg, :on_heredoc_beg @embexpr_balance += 1 when :on_embexpr_end, :on_heredoc_end @embexpr_balance -= 1 when :on_tstring_content if embexpr_balance == 0 token.value.split(/(?<=\n)/).each_with_index do |line, index| next if line.strip.empty? && line.end_with?("\n") next if !(dedent_next || index > 0) leading = line[/\A(\s*)\n?/, 1] next_dedent = 0 leading.each_char do |char| if char == "\t" next_dedent = next_dedent - (next_dedent % TAB_WIDTH) + TAB_WIDTH else next_dedent += 1 end end @dedent = [dedent, next_dedent].compact.min end end end @dedent_next = token.event == :on_tstring_content && embexpr_balance == 0 tokens << token end
# File yarp/lex_compat.rb, line 396 def to_a # If every line in the heredoc is blank, we still need to split up the # string content token into multiple tokens. if dedent.nil? results = [] embexpr_balance = 0 tokens.each do |token| case token.event when :on_embexpr_beg, :on_heredoc_beg embexpr_balance += 1 results << token when :on_embexpr_end, :on_heredoc_end embexpr_balance -= 1 results << token when :on_tstring_content if embexpr_balance == 0 lineno = token[0][0] column = token[0][1] token.value.split(/(?<=\n)/).each_with_index do |value, index| column = 0 if index > 0 results << Token.new([[lineno, column], :on_tstring_content, value, token.state]) lineno += 1 end else results << token end else results << token end end return results end # Otherwise, we're going to run through each token in the list and # insert on_ignored_sp tokens for the amount of dedent that we need to # perform. We also need to remove the dedent from the beginning of # each line of plain string content tokens. results = [] dedent_next = true embexpr_balance = 0 tokens.each do |token| # Notice that the structure of this conditional largely matches the # whitespace calculation we performed above. This is because # checking if the subsequent token needs to be dedented is common to # both the dedent calculation and the ignored_sp insertion. case token.event when :on_embexpr_beg embexpr_balance += 1 results << token when :on_embexpr_end embexpr_balance -= 1 results << token when :on_tstring_content if embexpr_balance == 0 # Here we're going to split the string on newlines, but maintain # the newlines in the resulting array. We'll do that with a look # behind assertion. splits = token.value.split(/(?<=\n)/) index = 0 while index < splits.length line = splits[index] lineno = token[0][0] + index column = token[0][1] # Blank lines do not count toward common leading whitespace # calculation and do not need to be dedented. if dedent_next || index > 0 column = 0 end # If the dedent is 0 and we're not supposed to dedent the next # line or this line doesn't start with whitespace, then we # should concatenate the rest of the string to match ripper. if dedent == 0 && (!dedent_next || !line.start_with?(/\s/)) line = splits[index..].join index = splits.length end # If we are supposed to dedent this line or if this is not the # first line of the string and this line isn't entirely blank, # then we need to insert an on_ignored_sp token and remove the # dedent from the beginning of the line. if (dedent > 0) && (dedent_next || index > 0) deleting = 0 deleted_chars = [] # Gather up all of the characters that we're going to # delete, stopping when you hit a character that would put # you over the dedent amount. line.each_char.with_index do |char, i| case char when "\r" if line.chars[i + 1] == "\n" break end when "\n" break when "\t" deleting = deleting - (deleting % TAB_WIDTH) + TAB_WIDTH else deleting += 1 end break if deleting > dedent deleted_chars << char end # If we have something to delete, then delete it from the # string and insert an on_ignored_sp token. if deleted_chars.any? ignored = deleted_chars.join line.delete_prefix!(ignored) results << Token.new([[lineno, 0], :on_ignored_sp, ignored, token[3]]) column = ignored.length end end results << Token.new([[lineno, column], token[1], line, token[3]]) unless line.empty? index += 1 end else results << token end else results << token end dedent_next = ((token.event == :on_tstring_content) || (token.event == :on_heredoc_end)) && embexpr_balance == 0 end results end