Fix issue where certain files would cause Errno::EPIPE exception to b…

…e raised (#9)

* Use Open3 instead of IO.popen for client_read

bin/console

@@ -0,0 +1,8 @@
+#!/usr/bin/env ruby
+# frozen_string_literal: true
+
+require 'bundler/setup'
+require 'henkei'
+
+require 'irb'
+IRB.start

lib/henkei.rb

@@ -11,6 +11,8 @@
 require 'socket'
 require 'stringio'
 
+require 'open3'
+
 # Read text and metadata from files and documents using Apache Tika toolkit
 class Henkei # rubocop:disable Metrics/ClassLength
   GEM_PATH = File.dirname(File.dirname(__FILE__))
@@ -224,11 +226,7 @@ def self.java_path
   # Internal helper for calling to Tika library directly
   #
   def self.client_read(type, data)
-    IO.popen tika_command(type), 'r+' do |io|
-      io.write data
-      io.close_write
-      io.read
-    end
+    Open3.capture2(tika_command(type), stdin_data: data).first
   end
   private_class_method :client_read
 

spec/henkei_spec.rb

@@ -40,6 +40,16 @@
       )
       expect(mimetype.extensions).to include 'docx'
     end
+
+    context 'when passing in the `pipe-error.png` test file' do
+      let(:data) { File.read 'spec/samples/pipe-error.png' }
+
+      it 'returns an empty result' do
+        text = Henkei.read :text, data
+
+        expect(text).to eq ''
+      end
+    end
   end
 
   describe '.new' do
@@ -129,6 +139,23 @@
     specify '#metadata reads metadata' do
       expect(henkei.metadata['Content-Type']).to eq %w[application/vnd.apple.pages application/vnd.apple.pages]
     end
+
+    context 'when passing in the `pipe-error.png` test file' do
+      let(:henkei) { Henkei.new 'spec/samples/pipe-error.png' }
+
+      it '#text returns an empty result' do
+        expect(henkei.text).to eq ''
+      end
+
+      it '#html returns an empty body' do
+        expect(henkei.html).to include '<body/>'
+        expect(henkei.html).to include '<meta name="tiff:ImageWidth" content="792"/>'
+      end
+
+      it '#mimetype returns an empty result' do
+        expect(henkei.mimetype.content_type).to eq 'image/png'
+      end
+    end
   end
 
   context 'initialized with a given URI' do