accumulo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Josh Elser <josh.el...@gmail.com>
Subject Re: Checkstyle Notes
Date Wed, 24 Dec 2014 20:25:18 GMT
At this point, I'd be in favor of just "fixing" the codebase. I spend so 
much time undo'ing formatting changes automatically made by Eclipse that 
are unrelated to my actual changes in an attempt to keep things minimized.

Merging will be more painful (hopefully Git is smart enough for most of 
it?), but maybe it's for the best? I'm not entirely sure.

Agreed in that we should avoid doing it twice if we're going to do it at 
all. Do you have some concrete changes in mind that we could poke around 
with?

Christopher wrote:
> Working on ACCUMULO-3451, what I found was that there's virtually no subset
> of rules I can apply that will not force some significant amount of our
> code to change in order to conform to those rules. it would be *much*
> easier if I just formatted the whole codebase first. I wouldn't want to do
> that twice in a short amount of time, though, if we're going to adopt
> google-styleguide any time soon.
>
> It seems that no matter what, enforcement of standards to prevent further
> quality decay, when we haven't been complying very well up to this point,
> is going to result in some disruption. The trick is getting the right
> balance/utility. Any thoughts?
>
>
> --
> Christopher L Tubbs II
> http://gravatar.com/ctubbsii
>
> On Tue, Dec 23, 2014 at 11:34 PM, Christopher<ctubbsii@apache.org>  wrote:
>
>> One problem I ran into, by the way, is that Eclipse does not actually
>> enforce the max line length. There's tons of bugs which have been open for
>> years, which prevent the formatter from wrapping in special circumstances.
>> The most common was simply that it ignored trailing punctuation. In other
>> words, if the only thing that would have wrapped was punctuation, it
>> doesn't wrap. This appears to be a competition between the rule about
>> putting parens/semicolons/braces on the same line and the rule for
>> wrapping. The same line appears to win.
>>
>> Checkstyle, on the other hand, gives a hard error on exceeding the line
>> length (which it should). This means that a file formatted with a formatter
>> set with a max line length of X may produce lines longer than X, and will
>> cause checkstyle to fail. The workaround is to manually insert
>> @formatter:off and @formatter:on lines so you can manually tweak to make
>> checkstyle happy without worrying about Eclipse reformatting. Currently,
>> these tags are not available in the google-styleguide. I have an open issue
>> about that, with a patch:
>> https://code.google.com/p/google-styleguide/issues/detail?id=27 . There's
>> also a few random bugs in the checkstyle implementation of the
>> google-styleguide (https://github.com/checkstyle/checkstyle/issues/533
>> and others), but these aren't a big deal, and in any case, only affect us
>> if we decide to adopt google-styleguide's rulesets.
>>
>>
>> --
>> Christopher L Tubbs II
>> http://gravatar.com/ctubbsii
>>
>> On Tue, Dec 23, 2014 at 1:24 PM, Christopher<ctubbsii@apache.org>  wrote:
>>
>>> Devs,
>>>
>>> So, I spent some time yesterday investigating the application of a
>>> CheckStyle configuration to apply to our builds.
>>>
>>> I started with the CheckStyle implementation[1] of the Google Style
>>> Guide[2], and a fully formatted Accumulo code base (according to our
>>> current Eclipse formatter standards[3]). I then configured the
>>> maven-checkstyle-plugin[4][5] to run during a build and built repeatedly,
>>> whittled down some of the strictness of the Google style and fixed some of
>>> our errors, to get at a minimal checkstyle configuration that might be able
>>> to add some additional checks during our build. I found at least one bug[6]
>>> doing this and evaluating the resulting warnings, and documented some of
>>> the common problems.
>>>
>>> Now, I'm not necessarily arguing for switching to adopting the Google
>>> Style Guide for our standards today (we can consider voting on that in a
>>> new thread). However, I think it might be beneficial to adopt the Google
>>> Style Guide, especially because that project has formatters for several
>>> standard IDEs, and because the latest version of CheckStyle has a ruleset
>>> which conforms to it, which enables the Maven tooling. So, if anybody is
>>> interesting in us doing that, I would certainly get behind it (and could
>>> help make it happen: reformatting, merging, and applying build tooling).
>>> One thing I like about the Google Style Guide, in particular, is that it
>>> doesn't just provide tools to enforce it, it also documents the standard
>>> with words and reasoning, so we have something to consult, even if you're
>>> not using an IDE or any of the tools.
>>>
>>> As a result of this, I am going to add something to start enforcing
>>> checks for some of the javadoc problems. We can add more checks later, or
>>> we can adopt the Google Style.
>>>
>>> In the meantime, here's a list of the most common style problems I
>>> observed (note: these aren't necessarily "bad", just non-conforming):
>>>
>>> Line length violations (we're currently using 160, but Google Style Guide
>>> allows 80 or 100, with 100 enforced in the tools by default; many of our
>>> lines exceed the 160).
>>> Escaped unicode in string literals (especially when they represent a
>>> printable character which should just be inserted directly, instead of
>>> encoded).
>>> Use of lower-case L for long literals (prefer 10L over 10l).
>>> Bad import order, use of wildcards.
>>> Missing empty line between license header and package declaration.
>>> Extra whitespace around generic parameters.
>>> Bad package/class/method/parameter/local variable naming conventions
>>> (like single character variables outside of loops, or use of
>>> underscore/non-standard camelCase).
>>> Missing optional braces for blocks (see a representatively bad example at
>>> CompactionInfo:lines 74-82).
>>> Keywords out of order from JLS recommendations (e.g. static public vs.
>>> public static).
>>> Overloaded methods aren't grouped together.
>>> Multiple statements per line, multiple variable declarations per line.
>>> Array brackets should be on type (String[] names;), not variable name
>>> (String names[];).
>>> Switch statements sometimes omit default case or fall through.
>>> Distance between local variable declaration and the first time it is used
>>> is sometimes too long (many lines).
>>> All uppercase abbreviations in names aren't avoided (LockID should be
>>> LockId).
>>> Operators should start the next line, when wrapping (like concatenation
>>> of two long string literals).
>>> Empty catch blocks (catch blocks should throw an AssertionError if it's
>>> supposed to be impossible to reach, or a comment about why it's safe to
>>> ignore).
>>> Some files have more than one top-level class. They should be in their
>>> own file.
>>> Braces not on same line as statement (like "} else" with "{" on next
>>> line).
>>>
>>> Missing javadocs on public methods.
>>> Missing paragraph markers(<p>) in javadoc paragraphs.
>>> Missing javadoc sentence descriptions.
>>> Javadoc tags out of order from standards.
>>> Missing html close tags in javadocs (<b>  but no</b>,<ul> 
but no</ul>)
>>> Use of<  and>  in javadoc instead of&lt; and&gt; or {@code ...}
>>>
>>> [1]:
>>> https://github.com/checkstyle/checkstyle/blob/master/google_checks.xml
>>> [2]: https://code.google.com/p/google-styleguide/
>>> [3]:
>>> https://github.com/apache/accumulo/blob/master/contrib/Eclipse-Accumulo-Codestyle.xml
>>> [4]:
>>> http://maven.apache.org/plugins/maven-checkstyle-plugin/check-mojo.html
>>> [5]: http://stackoverflow.com/a/27359107/196405
>>> [6]: https://issues.apache.org/jira/browse/ACCUMULO-3448
>>>
>>> --
>>> Christopher L Tubbs II
>>> http://gravatar.com/ctubbsii
>>>
>>
>

Mime
View raw message