# Experiment to tweak source based on what rubocop wanted me to.  I basically
# gave up and made the following exceptions.  Personally, I find the tool did
# find many interesting/questionable code Qs in the project but it is soooo
# opinionated in pretty unimportant ways (like who cares how I write my
# comments, or why can't I use kind_of? or why give me a fail for a less
# readable but more performant pattern).
#
# I can obviously tweak rules (like below) but I keep hearing of people pulling
# their hair out over all the unimportant rules; I question why this project is
# "all-in" on it's rules versus leaving most of the style rules as some preset
# profiles people can opt in to.
#
# So I am committing this and I will try and run this every now and then to
# see what it thinks about what I wrote but you will not see this as a
# requirement for commits since I think by the time I configured it to my
# liking I could have finished implementing this library.  Perhaps if I
# continue using this here and there I will end up with a profile I find
# acceptable?  I did find value but all the piddly shit just made me angry.
# I will reflect on it some more...

# This is a Java AWT call.  It is not what the tool thinks it is...which
# might be an issue for users interacting with Java from JRuby.
Lint/UselessSetterCall:
  Exclude:
    - 'lib/image_voodoo/awt.rb'

# This is just impossible for them to call properly.  I am doing something
# like rotate_impl and it is a bunch of reasonably simple math.  If you look
# for this transformation in a text book it will look like this code.  Putting
# a bunch of math within different methods would get the score down but
# it would not help readability. (addendum: I did try and break it apart more
# and never reached 15 but got closer.  I felt this rule is neat but the
# score really depends on what it is.
Metrics/AbcSize:
  Max: 16.5

# This is triggered for correct_orientation_impl.  I could put these transforms
# in a table and then execute lambdas for each orientation but this is much
# more clear.
Metrics/CyclomaticComplexity:
  Max: 9

# A bunch of fields in a hash are basically generated data. Correcting them
# for column seems much too pedantic.  I guess exclude is the right thing for
# an unusual file?
Metrics/LineLength:
  Max: 132
  Exclude:
    - 'lib/image_voodoo/metadata.rb'

# Metadata classes have data in them.  awt.rb is big and perhaps could be
# shrunk but it is not a hard file to navigate and the cleaving points are
# not super obvious.  These sorts of rules feel very arbitrary and if I have
# n highly related things and they do not fit into a more restrictive
# taxonomy why would the file being longer matter?  I do understand the
# motivation here but as a default rule this feels wrong to me.
Metrics/ClassLength:
  Max: 250

# I do not find this very useful.  There is no performance difference and
# sometimes I want to highlight this string does not involve interpolation.
# Other times it is not worth pointing out.
Style/StringLiterals:
  Enabled: false

# require 'english' is a step too far for $! which is so baked into my
# head I do not want to change :)
Style/SpecialGlobalVars:
  Enabled: false

# I am grouping math and using lack of whitespace for separation.
Layout/SpaceAroundOperators:
  Enabled: false

# I prefer tight assignment for opt args.
Layout/SpaceAroundEqualsInParameterDefault:
  Enabled: false

# Java methods which override or implement Java method names cannot be switched
# to snake case.  Do we really need this as a rule anyways?  I have never seen
# a Rubyist do this as a preferred style?
Naming/MethodName:
  Enabled: false

# bin/image_voodoo main options block.
Metrics/BlockLength:
  Exclude:
    - 'bin/image_voodoo'

# casecmp for this case seems like it is much less readable in a place where
# performance could never matter (a 3-4 char downcase before processing an
# image :) ).  This could end up being important somewhere but as a default
# on it feels weird since I find it less readable.
Performance/Casecmp:
  Exclude:
    - 'lib/image_voodoo/awt.rb'

# Hash rocket looks much more natural in a rakefile for its deps.
Style/HashSyntax:
  Exclude:
    - 'Rakefile'

# Forget it.  I do parallel assignment and you will have to peel it out of
# my cold dead hands.
Style/ParallelAssignment:
  Enabled: false

# FIXME: consider keywords for shapes.
# I might switch these to keyword args if I ever revisit shapes support.
# In general lots of params do suck and are hard to remember.  This library
# still is supposed to work in 1.8 but I can probably soon major rev this
# and switch over to keywords.
Metrics/ParameterLists:
  Exclude:
    - 'lib/image_voodoo/awt/shapes.rb'

# This is complaining that 'x' and 'y' are bad names because they are less
# than 3 chars.  Silly rule even if I appreciate what it is trying to
# accomplish. Interestingly, in image_voodoo I do use h for height and
# w for width.  Those probably could be written out but they are so common
# I am basically opting for short-hand.  This is something I rarely do except
# domain-driven code like this (x,y,h,w).
Naming/UncommunicativeMethodParamName:
  Enabled: false

# This represents one of those frustrating dogmatic rules where if I followed
# this advice I would then violate the column length rule.  I specifically
# use traditional if indented block in cases where if_mod doesn't fit.
# Gramatically, if_mod can be across two lines but I hate how that looks. The
# second solution of logical &&/|| change for a raise is alien to for all but
# assignment.
Style/IfUnlessModifier:
  Enabled: false

# What the hell.  Really telling people that encoding: utf-8 should not happen
# because there are no multi-byte utf-8 characters in it?  Why add this noise?
Style/Encoding:
  Enabled: false

# FIXME:
# This is an unknown error which I should figure out.  It must be some
# StandardError (which is what nothing explicit means) but I should actually
# be more explicit here.
Style/RescueStandardError:
  Exclude:
    - 'lib/image_voodoo.rb'


# This was complaining metadata_impl should use @metadata_impl instead of
# @metadata.  I disagree and am unsure why I need to follow this programs
# opinion on what a good name is?  What is commentable is if I change the
# variable name to @gorgon it still tells me I need to use @metadata_impl.
# So if I conditionally add any state in any method it will tell me I am
# memoizing and have to use same ivar as the name of the method?  Is it more
# sophisticated than that?  Should I have spent 2-3 minutes trying to figure
# out how this rule was implemented? :)
Naming/MemoizedInstanceVariableName:
  Enabled: false

# I have two modes of use for private.  If it is a small number of individual
# methods I am explicit and inline the private.  If it is a large number of
# methods especially when the name is fairly obvious that those will be private
# (like render_impl) then I use block private.  Not sure telling me I can only
# use a single style is very useful.
Style/AccessModifierDeclarations:
  Enabled: false

# I feel for the tiny amount of generation here a better name would not help
# me understand it better.  This is one of those naming rules I feel is
# probably helpful a minority of the time.  Also I have never been confused
# by what a heredoc text represents because of my heredoc terminator?
Naming/HeredocDelimiterNaming:
  Enabled: false