Support OpenOffice password decryption

[myplaceonline]

Support OpenOffice password decryption when encrypted with
AES-256-CBC (no padding) + PBKDF2/HmacSHA1 key and
SHA-256 password hash.

[myplaceonline]

Forgot to add that to use this feature, simply pass in the password in the options hash. For example, as shown in the test in the commit:

Roo::LibreOffice.new(File.join(TESTDIR, "encrypted-letmein.ods"), :password => "letmein")

[coveralls]

Coverage increased (+0.04%) when pulling 3d5adc3 on myplaceonline:master into 63996f0 on roo-rb:master.

[myplaceonline]

@tdeo Tried to add it, but got the following error:

  1) Error:
TestRoo#test_openoffice_encryption:
NoMethodError: undefined method `expect' for #<TestRoo:0x000000026ba9c0>
    /work/myplaceonline/src/src/roo/test/test_roo.rb:1961:in `test_openoffice_encryption'

Using:

      expect {
        Roo::LibreOffice.new(File.join(TESTDIR, "encrypted-letmein.ods"), :password => "badpassword")
      }.to raise_error ArgumentError, "Invalid password or other data error"

[simonoff]

I don't like the quality of this code.

[myplaceonline]

@simonoff Can you please elaborate which parts you don't like so that I can improve them?

[coveralls]

Coverage decreased (-0.23%) when pulling a22a413 on myplaceonline:master into 63996f0 on roo-rb:master.

[myplaceonline]

@simonoff The only obviously poor quality thing I could find was that there was superfluous file I/O. Pushed a new commit which optimizes this to stream from the zip file directly into the decryption cipher. Also added more comments and reformatted code to smaller column width and added a few more sanity checks.

[simonoff]

@myplaceonline it's additional 150 lines of code with 7 inherited if. Can you extract this lines into few methods?

[coveralls]

Coverage decreased (-0.06%) when pulling 966f3ce on myplaceonline:master into a0dd800 on roo-rb:master.

[myplaceonline]

@simonoff Extracted into some logical methods

[myplaceonline]

@simonoff I tried to merge in commit 85ca578 but it is currently failing tests:

$ git clone git@github.com:roo-rb/roo.git
...
$ cd roo
$ rake
...
  1) Failure:
TestBase#test_to_csv_with_separator [/tmp/roo/test/test_generic_spreadsheet.rb:124]:
--- expected
+++ actual
@@ -1,17 +1 @@
-";;;;;;
-;;;;;;
-;;;;;;
-;;;;;;
-1961-11-21;;;;;;
-;;;;;;
-;;;;;;
-;;\"thisisc8\";;;;\"thisisg8\"
-;;;;;;
-;;;;;;
-;;;;;;
-41;42;43;44;45;;
-;;;;;;
-;;;;;;
-;;43;44;45;;
-;;\"dreiundvierzig\";\"vierundvierzig\";\"fuenfundvierzig\";;
-"
+""
...

[myplaceonline]

@simonoff Test failure is caused by lines 136-137 in 85ca578#diff-e7ac28e488d98ee522c91f631315b536L136

The problem is that in the change to using a default parameter with sheet = default_sheet and removing sheet ||= default_sheet, it will break any callers that explicitly pass nil for sheet, as test_to_csv_with_separator does.

As in the case of test_to_csv_with_separator, this could be done by anyone that wants to pass a positional argument (such as the separator character in this case) and accept the defaults for previous arguments. Therefore, the multiple changes to use parameter defaults instead of ||= in commit 85ca578 effectively break the APIs for such callers.

If it is decided to continue with commit 85ca578, then the test can be fixed with:

    assert_equal expected_csv_with_semicolons,@oo.to_csv(nil, @oo.default_sheet, ';')

I will open an issue for this so it's documented as a separate issue...

[myplaceonline]

Opened Issue #159

[myplaceonline]

@simonoff I've merged in all the latest commits, please review when you have a chance (noting that the Travis CI build errors are caused by issue #159 which is separate).

[ebertech]

aren't there laws against using camelCasing in Ruby (TM) programs? getCipherKey? decryptIfNecessary?

[myplaceonline]

@ebertech Updated, thanks

[ebertech]

@myplaceonline you're welcome! yeah opensource!

[myplaceonline]

@simonoff Issue #159 fixed, changes merged, CI build passed, ready to merge

[coveralls]

Changes Unknown when pulling 20d77db on myplaceonline:master into * on roo-rb:master*.