From c939c0121332210369be913c3fd3f0ec13fa2a87 Mon Sep 17 00:00:00 2001
From: chopraanmol1 <chopraanmol1@gmail.com>
Date: Tue, 18 Sep 2018 18:14:04 +0530
Subject: [PATCH] Introduce GC'able cache for Roo::Excelx::Extractor#doc

Memory Profiling script:
```
MemoryProfiler.report{Roo::Excelx.new('test/files/Bibelbund.xlsx').tap{|x|(2..x.last_row).each{|i| x.row(i)}}}.pretty_print

```

Benchmark Script:
```
puts Benchmark.measure{ Roo::Excelx.new('test/files/Bibelbund.xlsx').tap{|x|(2..x.last_row).each{|i| x.row(i)}} }
```


Result on Master:
```
Total allocated: 37131810 bytes (517026 objects)
Total retained:  5562913 bytes (103010 objects)

allocated memory by gem
-----------------------------------
  19288066  roo/lib
  11049821  nokogiri-1.8.4
   6792403  rubyzip-1.2.2
      1304  tmpdir
       216  other

retained memory by gem
-----------------------------------
   5560934  roo/lib
       782  rubyzip-1.2.2
       725  nokogiri-1.8.4
       296  tmpdir
       176  other

  0.720000   0.020000   0.740000 (  0.733750)
```

Result after patch:
```
Total allocated: 34561842 bytes (504998 objects)
Total retained:  5563553 bytes (103026 objects)

allocated memory by gem
-----------------------------------
  19254338  roo/lib
   8513101  nokogiri-1.8.4
   6792403  rubyzip-1.2.2
      1304  tmpdir
       320  weakref
       216  other
       160  ref-2.0.0

retained memory by gem
-----------------------------------
   5561094  roo/lib
       782  rubyzip-1.2.2
       725  nokogiri-1.8.4
       320  weakref
       296  tmpdir
       176  other
       160  ref-2.0.0

  0.610000   0.010000   0.620000 (  0.618642)
```


Note:
Ruby does have native implementation of WeakRef. But I've choosed to go with 'ref' gem for this following reason:
	https://github.com/ruby-concurrency/ref#problems-with-weakref
---
 lib/roo/excelx/extractor.rb | 24 +++++++++++++++++++++---
 roo.gemspec                 |  1 +
 2 files changed, 22 insertions(+), 3 deletions(-)

diff --git a/lib/roo/excelx/extractor.rb b/lib/roo/excelx/extractor.rb
index 0b3006a1..adc83f39 100755
--- a/lib/roo/excelx/extractor.rb
+++ b/lib/roo/excelx/extractor.rb
@@ -1,9 +1,10 @@
 # frozen_string_literal: true
 
+require "ref"
+
 module Roo
   class Excelx
     class Extractor
-
       COMMON_STRINGS = {
         t: "t",
         r: "r",
@@ -21,14 +22,31 @@ def initialize(path, options = {})
       private
 
       def doc
-        raise FileNotFound, "#{@path} file not found" unless doc_exists?
+        instance_cache(:@doc) do
+          raise FileNotFound, "#{@path} file not found" unless doc_exists?
 
-        ::Roo::Utils.load_xml(@path).remove_namespaces!
+          ::Roo::Utils.load_xml(@path).remove_namespaces!
+        end
       end
 
       def doc_exists?
         @path && File.exist?(@path)
       end
+
+      def instance_cache(key)
+        object = nil
+
+        if ref = instance_variable_get(key)
+          object = ref.object
+        end
+
+        unless object
+          object = yield
+          instance_variable_set(key, Ref::WeakReference.new(object))
+        end
+
+        object
+      end
     end
   end
 end
diff --git a/roo.gemspec b/roo.gemspec
index 62dbb82a..75fe4fda 100644
--- a/roo.gemspec
+++ b/roo.gemspec
@@ -18,6 +18,7 @@ Gem::Specification.new do |spec|
   spec.require_paths = ['lib']
 
   spec.add_dependency 'nokogiri', '~> 1'
+  spec.add_dependency 'ref', '~> 2'
   spec.add_dependency 'rubyzip', '>= 1.2.1', '< 2.0.0'
 
   spec.add_development_dependency 'rake', '~> 10.1'