[
https://issues.apache.org/jira/browse/MAHOUT-913?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13165449#comment-13165449
]
Dmitriy Lyubimov commented on MAHOUT-913:
-----------------------------------------
1) FWIW i always used 1d too.
The reasoning is since there's no substitute for specifying 1l and 1f and it
just so happens that 1.0 by default transforms into 1d but not 1f or 1l.
Since you can't avoid using 1l and 1f where you need them (and you do need them
esp. with in case of longs), there's no reason avoiding using 1d. In fact,
using 1d is more explicit and conceptually coherent.
single precision arithmetic case may be less evident these days (but still
valid in case you want e.g. to save space with matrix stuff) but longs have a
much stronger case. E.g. it should be evident that long a = (1<<31) and long a
= (1l<<31) produce completely different results and there's no immediate
substitute for 1l notation in the second case using default transformation rule
such as 1.0.
Am not arguing about need for uniform approach, just that 1d makes more sense
to me.
> Style changes / discussion
> --------------------------
>
> Key: MAHOUT-913
> URL: https://issues.apache.org/jira/browse/MAHOUT-913
> Project: Mahout
> Issue Type: Improvement
> Affects Versions: 0.5
> Reporter: Sean Owen
> Assignee: Sean Owen
> Priority: Minor
> Fix For: 0.6
>
> Attachments: MAHOUT-913.patch
>
>
> Guys I've still been seeing code committed that doesn't match standard Java
> style or a reasonable policy I can imagine. I wanted to talk about it since
> I've just been silently changing it and that is not ideal.
> This should be easy to get right, as automated tools exist to check and fix
> this. I recommend IntelliJ's free Community edition. Flip on even basic
> inspections. A hundred things will jump out (that are already jumping out at
> me). Most are automatically fixable.
> I think that standardized, readable code invites attention, work and care: it
> feels like something you want to improve, and don't want to hack up.
> I think it helps attract committers. Strong engineering organizations
> wouldn't let basic style problems in the codebase, just by using automated
> checks. Code reviews don't begin otherwise, and then reviews focus on real
> issues like design. We can make a basic effort to approach that level of
> quality. Otherwise, people who are used to a higher standard won't be
> inclined to participate in the project, and will just fork.
> I think it's a prerequisite to fixing real design issues, TODOs, correctness
> problems (cloning for instance), and refactorings. This code is not near that
> point, and won't get there at this rate.
> Personally it makes we want to only support anything I've written, and write
> any "next generation" recommender system in a new and separate venture. And
> I'm a friendly, and maybe not the only one! So would be great to keep some
> focus on quality and design.
> Here's a patch showing all the changes I've picked up and made with the IDE
> -- *just* basic style issues, and just since the last 2 weeks. The issues
> are, among others:
> ⁃ Empty javadoc
> ⁃ Redundant javadoc ("@param foo the foo")
> ⁃ Missing copyright headers
> ⁃ Copyright headers not at top of file (sometimes after imports!)
> ⁃ Very long lines (>> 120 chars)
> ⁃ "throws Exception" not on main() or test method
> ⁃ "transient" fields -- should never be used for us
> ⁃ Missing @Override
> ⁃ Using new Random()
> ⁃ Redundant boolean expressions like "foo == true"
> ⁃ Unused variables and parameters
> ⁃ Unused imports
> ⁃ Loops and conditionals without braces
> ⁃ Weird literals ("1d")
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
