@passport-next/eslint-config-passport-next

Forked from https://github.com/brettz9/eslint-config-ash-nazg

Status:

A ruthless--though not arbitrary--degree of control for your kingdom!

One ring to rule them all!

The rules

See index.js (and node.js for ash-nazg/node rules) for the rules we explicitly include (and see sauron.js for the even stricter ash-nazg/sauron rules or great-eye (or great-eye-node) for still stricter rules though which are probably best not used).

(See explicitly-unused.js for the core and extended rules we don't include (rationale for non-inclusion below).)

• Main rules (also search for "error" within https://github.com/eslint/eslint/blob/master/conf/eslint-recommended.js) for the eslint:recommended rules we inherit (though see below for our two modifications to these).
• "standard" rules we inherit (though see below for our handful of modifications).
• mysticatea/eslint-plugin for a number of added rules (though with a few items disabled and enabled as per below), including all of @mysticatea/eslint-comments/recommended.
• eslint-plugin-no-use-extend-native for one added rule.
• Recommended Unicorn rules (a few items disabled and enabled as per below)
• eslint-plugin-array-func for further rules

I've focused below on deviations because the original sites tend to already articulate the usefulness of the rules I have incorporated for these plugins/configs. This is for tracking design choices and not disparaging the utility of non-included rules.

Other bundled configs

The ash-nazg/node config expands upon the regular ash-nazg rules to add rules specific to Node environments. Specifically, plugin:node/recommended-module has been adopted for now along with some stylistic choices.

The ash-nazg/sauron config expands upon the regular ash-nazg rules to indicate what are generally best practices but are less likely to be due to an error and may possibly also require a high (and possibly tyrannical) degree of refactoring for existing projects. See below for the rationales for inclusion.

The ash-nazg/great-eye and ash-nazg/great-eye-node configs expands on ash-nazg/sauron (and ash-nazg/sauron-node) to include rules which enforce good practices, but which are so cumbersome and may flag too much legitimate code that I personally won't regularly use them. Still, I like to track them here, including in the event that their config changes to make them less all-encompassing.

The ash-nazg/sauron-node config incorporates both ash-nazg/node and ash-nazg/sauron. It adds specific rules of its own which may be unduly strict for ash-nazg/node.

Comparison to other "standards"

eslint-config-ash-nazg is a collection of some excellent preexisting rules other projects have created, curated with the intent of attaining productive strictness.

"standard" (config) and eslint:recommended have brought a welcome degree of pseudo-standardization for code styling and error checking.

However, they do not impose as high of a degree of control as one may find desirable to enforce common, sane styling practices, catch bad practices and hard-to-read patterns, or in some cases even catch errors.

Moreover, standard enforces a few styling rules which are contrary to regular JavaScript practice (e.g., its enforced absence of semi-colons).

ash-nazg aims to offer defaults which adhere to norms, build upon productive restrictions, and in a few cases, offer greater latitude where some constraints are unduly confining.

To see which rules from an extended config ended up enabled (we are typically inheriting "recommended" configs), see /inherited-rules/implicitly-included.

To see which rules from the extended config we disabled, see our relevant config(s) (e.g., index.js).

To see which rules from the plugin that each extended config derives from (i.e., the non-recommended rules of a plug-in), see explicitly-unused.js for ones we have consciously not used and see our relevant config(s) (e.g., index.js) for the ones we did add (alongside any recommended).

The unused folder is used to capture any (non-recommended) rules which are not explicitly either enabled or within explicitly-unused.js (as might be found during an update of our config dependencies), but it is currently, and should hopefully remain, empty, as we wish to be consciously aware of all rules from inherited projects and whether we wished to include them. As far as new rules added to recommended, we can see these within diffs of inherited-rules/implicitly-included files (built during development by npm run compare).

Rationales for inclusion of dependent plug-ins

Besides incorporating more from ESLint core, we also add rules from a few other (peer) dependencies.

• eslint-plugin-compat - Though its list of browsers to support is not added by our code, it is a basic enough of a need to remind one of its need to be present (and to avoid needing to add the rule for it).
• eslint-plugin-import - Besides being peer dependencies of standard, this has some additional useful rules we apply.
• eslint-plugin-jsdoc - provides a number of helpful rules for catching bad or incomplete JSDoc during development. Note that we do not impose jsdoc/require-jsdoc except as a warning in ash-nazg/sauron mode, so all of the eslint-plugin-jsdoc rules should only apply when JSDoc is used. We also avoid some of eslint-plugin-jsdoc stricter checking requirements for descriptions.
• eslint-plugin-markdown - Another basic use case is ensuring code snippets are linted, which is why this is expected. You may find its discussion of overrides helpful for tweaking rules for Markdown (and you may also wish to use overrides, for the eslint-plugin-jsdoc rule check-examples if you wish to lint your JSDoc examples according to different standards). You must pass the file extension within the --ext flag, e.g., --ext .md,.js to get this to take effect.
• eslint-plugin-html - One basic use case is ensuring HTML is linted, which is why this is expected. You may wish to use overrides for HTML-specific rules or enable the plugin's own rules. You must pass the file extension within the --ext flag, e.g., --ext .html,.htm,.js to get this to take effect.
• eslint-plugin-promise - Besides being peer dependencies of standard, this has some additional useful rules we apply.

Deviations

Deviation from eslint:recommended

In comparison to eslint:recommended, ash-nazg only adds restrictions with the exceptions of:

1. Loosening the ESLint < 6 requirement of no-console into a mere warning and only in the stricter ash-nazg/sauron configuration. This is for the reason that console logging is too useful for debugging (and sometimes for reporting progress) to have to disable it at every turn and to distract one from actual errors.
2. Change no-empty so that allowEmptyCatch is true. There are enough cases where one legitimately needs to suppress errors.

Rationale for overriding standard rules

1. semi - Even if not strictly required, semi-colons are conventional in JavaScript and help denote the end of statements (as opposed to the end of a line which may continue).
2. indent - While it may take some getting used to, 2 spaces does allow more in one's field of vision. However, changed to use outerIIFEBody for avoiding indents with the IIFE body, as this often minimizes indent level for much or all of the whole file.
3. object-property-newline - Properties on the same line can be very convenient, including stacking for space to avoid max length (though without stacking the height too high).
4. one-var - While I normally favor enforcing conventions, this one seems to me to be of little consequence. It also prevents grouping like items together. I might favor an option to require separate lines for variable declarations, but for uninitialized ones, adding to the same line is convenient, especially for single-letter variables. I would like a rule to have declarations as close as possible to being just above any used scopes (for let and const).
5. object-curly-spacing - Not sure why standard switched from the default here.

Rationale for changing required rules' configuration away from ESLint defaults

While these are not part of standard or eslint-recommended, I've noted here why we have deviated from the defaults set by ESLint for those applying the rule.

• function-paren-newline - The default multiline can get too long whereas "consistent" can be clean and short.
• multiline-ternary - Inline ternary can be very readable when not spanning lines
• no-empty - Empty catches occur frequently enough to justify allowEmptyCatch
• no-mixed-requires - Grouping is more organized, while calls are compellingly convenient
• no-restricted-globals - Use example defaults for event (and fdescribe) as convenient
• no-restricted-properties - Use example default of restricting __defineGetter__ in favor of Object.defineProperty as convenient
• no-restricted-syntax - Set to eliminate any use of instanceof (as it is not cross-frame safe, except if using Symbol.hasInstance)
• no-shadow - It is better practice not to confuse by using globals! I didn't feel the examples for allowing necessitated their use. I added "parent", "top", "open", "close", "stop", "blur", "status", "name", "closed", and "start" to allow as frequent local variables technically shadowing globals (start for QUnit env).
• object-shorthand - Enabling avoidExplicitReturnArrows as methods without this that this rule grabs (which have their own meaning with arrow functions) are more succinctly expressed in shorthand, and shorthand also conduces better to events where this refers to the element (unless one needs the outer scope this). The option avoidExplicitReturnArrows does not avoid if the functions have this (though this isn't currently documented); this is probably good, though use of this may signal an error if within a method
• prefer-destructuring - Did not set array to true due to its problems with direct access of large numbers (requiring many commas) and non-iterable ‘array-like’ objects.
• quote-props - Changed to "as-needed" as properties more verbose and uglier with quoting
• import/order - Enforcing "builtin", "external", "internal" and then ["parent", "sibling", "index"] in any order as these may be project-specific.

Rationale for not including some rules

The following rules are not included within eslint:recommended or standard, nor have we chosen to include them in any of the ash-nazg rules.

Note: italicized items refer to features I might enable if the described option could become available (or upon further review). I might also tweak some standard rules further which I have not had time to examine (but it would probably be toward the stricter rather than looser as I have been happy with it thus far).

• array-bracket-newline - Switched to "consistent" over "always" as sometimes nice to be a little condensed, especially when representing HTML as JSON within Jamilih (or JsonML) templates.
• array-element-newline - While the "consistent" option would be nice, it doesn't work well to keep up with a max width and the desire to avoid excessive height [\n a, b, \n c, d\n]
• arrow-body-style - With as-needed and requireReturnForObjectLiteral, this seems reasonable, but too often in debugging, one needs to add brackets to do any logging.
• radix - We're only dealing with ES5+ environments, so the radix is redundant for base 10.
• func-names - Too prohibitive, though if applied to methods only, it may be useful (though with object shorthand, less necessary)
• func-style - Declarations are simpler so appealing also. If enabling, would also want allowArrowFunctions.
• id-blacklist - Can be helpful but a little tyrannical
• id-length - Can be helpful (especially a minimum) but a little tyrannical
• id-match - Too project-specific
• init-declarations - Nice with "always" and ignoreForLoopInit, but cumbersome and seemingly wasteful at times
• line-comment-position - Too inflexible to enforce either way
• lines-around-comment - Irksome to me
• lines-between-class-members - Don't feel any need for it
• max-classes-per-file - A bit tyrannical
• max-depth - Sounds good but tyrannical
• max-lines-per-function - A bit tyrannical
• max-lines - A bit tyrannical
• max-params - Can be troublesome when one is forced to abide by some API
• multiline-comment-style - Would be nice if allowed multiline "starred-block" OR "bare-block" given some one may wish as JSDoc-style and others not
• newline-after-var - Not very flexible
• newline-before-return - Not very flexible
• newline-per-chained-call - Not flexible in practice
• no-continue - Can be convenient
• no-invalid-this - Sounds good but not useful for element event handlers/jQuery
• no-multi-assign - Sounds good, but can be burdensome
• no-negated-condition - Is generally simpler, but it can be annoying if one wishes to get a much smaller condition body visible at the top.
• no-nested-ternary - Nested ternaries can be helpful to avoid clutter of duplicated assignment code
• no-param-reassign - Can be helpful, but not convenient, including when making defaults against more than undefined (e.g., null)
• no-restricted-imports - Project-specific
• no-restricted-modules - Project-specific
• no-restricted-properties - Might expand the list
• no-restricted-syntax - Project-specific unless some forms exist which should be deprecated
• no-ternary - Not useful
• no-undefined - undefined is ok for ES6 modules and strict code, so using no-global-assign and no-shadow-restricted-names instead
• no-underscore-dangle - Too restrictive
• no-useless-concat - Too restrictive when one has certain formatting one wishes to draw out
• object-curly-newline - Doesn't allow let f = {foo () { dosomething(); }};
• one-var-declaration-per-line - Sounds good, but too cumbersome for small integer or boolean inits, and the indented next lines are not as immediately clear that they belong to the declaration.
• padding-line-between-statements - Might revisit
• prefer-arrow-callback - Not compelling
• prefer-template - Sounds good, but too cumbersome in practice
• sort-imports - Would be useful with "warn" if could sort by multiple/single type and sort members while avoiding alphabetical sorting across imports which seems too rigid
• sort-keys - Too cumbersome
• sort-vars - Too cumbersome

Rationale for disabled import rules

• import/no-restricted-paths - Project-specific
• import/no-internal-modules - Don't see any advantage
• import/no-cycle - Don't see a problem with cyclic imports in ESM
• import/no-nodejs-modules - Useful for some projects, but not all including even client-side (though does offer allow option)
• import/exports-last - Has some reason for being, but nicer to see with declaration that the object is being exported. To find all, just search for "export"
• import/no-namespace - While can be more efficient to import only what one needs, having a namespace can also avoid confusion.
• import/prefer-default-export - Could even be a bad practice as ugly to import with non-defaults added later
• import/no-default-export - Has some basis, but defaults are admittedly convenient. Let's not be so opinionated.
• import/no-named-export - Could even be a bad practice as ugly to import with non-defaults added later
• import/no-relative-parent-imports - Appealing in some ways, but too rigid for a broad standard
• import/dynamic-import-chunkname - Might revisit for warnings, but probably too specific for a broad standard
• import/max-dependencies - Too constraining
• import/no-unassigned-import - Could be useful with allow option, but that would be project-specific, and unassigned imports are needed for polyfills
• import/group-exports - Too rigid as with exports-last.

Rationale for disabled Node and Promise rules

node/prefer-promises/dns and node/prefer-promises/fs are good, but a bit early with Node 12.

node/no-callback-literal duplicates standard/no-callback-literal (and is not recommended anyhow).

promise/no-native is disabled as promises are essential--even, it appears, to Dark Lords.

promise/param-names can be too tyrannical in some cases.

Rationale for suppressing some eslint-plugin-jsdoc rules

• jsdoc/no-types - Types have utility in jsdoc unless using TypeScript
• jsdoc/newline-after-description (recommended) - I can see its draw, but seems too pedantic to me for documentation.
• jsdoc/no-undefined-types (recommended) - I'd like something like this, but since it isn't aware of all of one's @typedefs, etc., it is too restrictive for me at this time.
• jsdoc/require-description-complete-sentence seems like a good idea, as English mistakes can be jarring as with bad styling, but this didn't seem to work too well. I use jsdoc/match-description with a special value to capture a subset of this.
• jsdoc/require-hyphen-before-param-description - I can see its draw, but seems too pedantic to me for documentation.

Rationale for disabling some unicorn rules

(The following are recommended rules unless otherwise noted.)

• catch-error-name - It can actually be useful to use different error names to indicate what time of error may be expected.
• explicit-length-check - Seems wasteful.
• filename-case - Looks potentially useful with camelCase.
• throw-new-error - Potentially confining.
• no-unreadable-array-destructuring - Better to use this than multiple lines
• import-index - While understandable, seems may cause more trouble in making it harder to find references to index.
• no-unused-properties - While no doubt useful, it won't catch all cases, sounds computationally expensive, and may better be done with TypeScript
• no-keyword-prefix - See no need.

Rationale for including some Unicorn rules which are disabled in plugin:unicorn/recommended

• unicorn/custom-error-definition - Does not seem confining.
• unicorn/no-unsafe-regex - Seems like something to watch for, though I may revisit if this reports too many common expressions.

Rationale for including eslint-comments rules which are not in plugin:eslint-comments/recommended

• eslint-comments/no-unused-disable - Might be indicative of an error and inefficient in any case

Rationale for changing eslint-comments rules from default recommended

eslint-comments/disable-enable-pair - If at top, behavior is clear, and no need to reenable within doc

Rationale for not including some plugin:@mysticatea/es2019 rules

• @mysticatea/arrow-parens - Covered by other rules
• @mysticatea/no-instanceof-array - Covered by our blocking of all instanceof
• @mysticatea/no-instanceof-wrapper - Covered by our blocking of all instanceof
• @mysticatea/prefer-for-of - I prefer array extras for easier reuse, currying, etc. than for-of
• @mysticatea/eslint-comments/no-restricted-disable - See no need
• @mysticatea/eslint-comments/no-use - See no need
• @mysticatea/eslint-plugin/* - Project more generic than ESLint plugins
• @mysticatea/ts/* - Not supporting TypeScript
• @mysticatea/vue/* - Project more generic than Vue.js
• @mysticatea/prettier - Not interested in imposing prettier
• @mysticatea/no-use-ignored-vars - Relies on a regex (for pseudo-privates) which can be useful

Rationale for not including some array-func rules

• array-func/prefer-array-from - While it may benefit performance, it is more sleek to use the spread operator. Would like to know how much it impacts performance before enabling.

Rationale for not including some sonarjs rules

(All sonarjs rules are currently "recommended" rules as well.)

• max-switch-cases - Sounds too arbitrary.
• no-collapsible-if - Sometimes more logically clear or made in preparation for future expansion
• no-duplicate-string - Often used in test files and repeated in fairly minor instances or instances repeated but spread out within a large file
• no-identical-functions - Often used in test files and repeated by fairly minor functions or functions repeated but spread out within a large file
• no-small-switch - Too useful to start a pattern to which one intends to add later.
• prefer-immediate-return - Can be useful for documenting even if method name should be descriptive, especially if there are different returns with subtly different results

Rationale for including eslint-plugin-jsdoc rules which are not in plugin:jsdoc/recommended

• check-examples - If examples are present, they ought to follow one's standards, including if overrides are in place to loosen/tighten. Set to match any Markdown rules (.md extension) by default and excludes any example beginning with a backtick.
• check-syntax - Following jsdoc, not Closure syntax
• match-description - Cleaner to see complete sentences which its default allows.
• require-returns-check - If the return value doesn't match, there may be a problem.

Rationale for only including some rules within ash-nazg/sauron

These are good practices, but cumbersome, not as familiar to developers, prohibitive during ongoing debugging or conversion of existing projects, etc. But perhaps useful for a new project which can pay closer attention to standards without the undue burden of having to refactor lots of code (which may not all be under one's control).

• class-methods-use-this - Good for new code (though even here may need disabling). May break existing APIs.
• consistent-this - A good practice, but sometimes, especially with jQuery or DOM event handlers, and large functions where it can be difficult to track what a generic that is in reference to, this may be more clearly named as the object
• default-case - Refactoring code to know whether to throw or silently ignore requires time.
• max-len - May require a lot of refactoring.
• no-alert - No quick easy replacement yet with poor dialog support.
• no-console - Useful for debugging (and sometimes for reporting progress) to have to disable it at every turn and to distract one from actual errors, even for production.
• no-empty-function - Cumbersome to add comments within every no-op.
• no-shadow - Can catch errors, but also some work to refactor (See also section "Rationale for changing required rules' configuration away from defaults".)
• prefer-numeric-literals - Good but some work.
• require-unicode-regexp - Good, but some work to fix all.
• vars-on-top - Not needed for let/const, and if overriding, this is cumbersome, despite being useful
• import/unambiguous - A good practice, and one which overrides can override (e.g., for polyfills or simple HTML tests), but cumbersome to label all files
• promise/prefer-await-to-callbacks - Sometimes useful, but callbacks may be used within repeating events
• unicorn/no-fn-reference-in-iterator - May be cumbersome though does catch potential problems
• jsdoc/require-returns (recommended) - Put in ash-nazg/sauron as it is more than just a consistent styling convention, and it is not impossible to follow, but a bit difficult. Added forceRequireReturn option to ensure return type considered even if void/undefined.
• jsdoc/require-jsdoc (recommended) - Imposes a heavy burden on preexisting large projects (added as "error" in great-eye.js)

The forceRequireReturn setting was also applied therein as it may be cumbersome to add to all returns or not favored as a requirement in all projects though it does note that a method's return was considered even if undefined.

The preferredTypes setting was enabled here as it can be cumbersome for projects to specify all child types.

Rationale for only including some rules within ash-nazg/great-eye

• capitalized-comments - May be very prevalent; reports with commented out code
• complexity - A generally good practice, but can be work to refactor, and sometimes impractical to enforce.
• max-statements-per-line - Tyrannical when prohibiting single-line if (...) { continue; }; might revisit if allowed for control statements
• max-statements - A bit tyrannical even if good for clear code
• no-magic-numbers - Very helpful for clear code, but cumbersome, and sometimes very cumbersome.
• no-plusplus - Would be nice if there were an option to allow if not combined inline with other expressions
• no-warning-comments - Good to catch to-dos, but better to search or parse code as a separate process rather than polluting one's ESLint warnings--some to-dos are ok to be left for the long term
• import/no-unused-modules - Useful (for missingExports at least), but doesn't catch dynamic imports and reports for other deliberately non-modular scripts
• jsdoc/check-alignment (recommended) - A pretty good practice, but not that important.
• jsdoc/check-indentation - A pretty good practice, but that that important.
• jsdoc/require-description - Though a very good practices, this is difficult for large code bases to implement and even cumbersome for new ones to have to follow for lesser parameters, so that's why they are not even in ash-nazg/sauron, though if you feel you can succeed at imposing such rigor for your hordes of developer minions, by all means, have a hand at it...
• jsdoc/require-param-description (recommended) - See jsdoc/require-description.
• jsdoc/require-returns-description (recommended) - See jsdoc/require-description.
• jsdoc/require-example - See jsdoc/require-description.
• sonarjs/cognitive-complexity - As with complexity perhaps (though may add to sauron if demonstrates not to be too oppressive)
• unicorn/prevent-abbreviations - Very cumbersome for frequent conventions such as e for event

The preferredTypes setting was enabled here for integer/float as it can be cumbersome for projects to distinguish.

Rationale for including rules that might not seem necessary

• no-implicit-globals - Included despite not applying to modules, in case overriding.
• strict - Included despite not being needed for modules, in case overriding.

Deprecated rule non-inclusion

Deprecated and removed items from ESLint are also naturally not included.

Contributing

While I will admit to being opinionated, and one may need to disable some rules (or possibly add a few ones mentioned in my non-inclusion sections), feel free to file issues if you really feel there are compelling reasons for different defaults. But again, I have to add caution that ring bearers can be picky about giving up their preh-shus...

To-dos

• Waiting: Add eslint-plugin-no-constructor-bind when class properties start landing
• Add own rules:
  1. tweak for no-new-object
  2. allow npm_config_* with no-process-env
  3. to-do query tool/extractor
• Review https://github.com/dustinspecker/awesome-eslint#practices
• See if https://github.com/sarbbottam/eslint-find-rules may eliminate need for our comparison code