brilliant test by @leomao10

lib/unix_utils.rb

@@ -192,12 +192,13 @@ def self.dos2unix(infile)
     end
   end
 
+  # POSIX sed, whether it's provided by sed or gsed
   def self.sed(infile, *expr)
     infile = ::File.expand_path infile
     outfile = tmp_path infile
-    bin = available?('gsed') ? 'gsed' : 'sed'
-    argv = [ bin, expr.map { |e| ['-e', e] } ].flatten
-    spawn argv, :read_from => infile, :write_to => outfile
+    bin = available?('sed') ? 'sed' : ['gsed', '--posix']
+    argv = [ bin, expr.map { |e| ['-e', e] }, infile ].flatten
+    spawn argv, :write_to => outfile
     outfile
   end
 

test/regression/test_sed_large_file.rb

@@ -0,0 +1,32 @@
+# encoding: UTF-8
+require 'helper'
+require 'tempfile'
+
+describe "handle large file with sed" do
+  let(:normal_line_no)  { 124000 }
+  let(:invalid_line_no) { 10 }
+  let(:input_file)      { Tempfile.open('test', '/tmp') }
+
+  before do
+    begin
+      (1..normal_line_no).each do |i|
+        input_file.puts "\"psmith01\",\"CLASS2B\",\"Peter Smith 1\",\"YEAR2\",\"1\",\"N\",\"ADVANCED\",\"STAFF\",\"1\",\"Y\",\"Y\",\"psmith01\",\"CLASS2B\",\"Peter Smith 1\",\"YEAR2\",\"1\",\"N\",\"ADVANCED\",\"STAFF\",\"1\",\"Y\",\"Y\",\"psmith01\",\"CLASS2B\",\"Peter Smith 1\",\"YEAR2\",\"1\",\"N\",\"ADVANCED\",\"STAFF\",\"1\",\"Y\",\"Y\",\"psmith01\",\"CLASS2B\",\"Peter Smith 1\",\"YEAR2\",\"1\",\"N\",\"ADVANCED\",\"STAFF\",\"1\",\"Y\",\"Y\""
+      end
+      (1..invalid_line_no).each do |i|
+        input_file.puts "@!!@"
+      end
+      input_file.flush
+    ensure
+      input_file.close
+    end
+  end
+
+  it "should have 124000 lines in output file" do
+    output_path = UnixUtils.sed(input_file.path, ':a', "1,#{invalid_line_no}!{P;N;D;};N;ba")
+    UnixUtils.wc(output_path).first.must_equal normal_line_no
+  end
+
+  after do
+    input_file.unlink
+  end
+end