Change Log

This changelog is loose. Versions are not semantic, they are incremental. Splint is not meant to be infrastructure, so don't rely on it like infrastructure; it is a helpful development tool.

Unreleased

v1.10.0

The big feature here is adding support to run splint without specifying paths. Now Splint will read the deps.edn or project.clj file in the current directory and check the paths in the base definition as well as :dev and :test aliases/profiles if no path argument is given. Splint still doesn't support specifying paths in .splint.edn, nor does it allow for using paths from a project file as well as additional paths when called, but those are planned.

The second big change is moving from the old DSL to the new pangloss/pattern inspired DSL. More flexible, more usable, overall better.

The third is adding performance rules. These are off by default and are designed for those who want to pinch their cycles. Some affect common code (get-in) and some are much more rare (satisfies), but they're all designed to help you be mindful of slower paths.

Breaking

• Moved spat.parser to splint.parser.
• Moved spat.pattern to splint.pattern. RIP spat, you treated me well for 9 months, but keeping spat and splint separate is no longer helpful.
• Switched to the new pattern system, updated all rules.

New Rules

• performance/assoc-many: Prefer (-> m (assoc :k1 1) (assoc :k2 2)) over (assoc m :k1 1 :k2 2).
• performance/avoid-satisfies: Do not use clojure.core/satisfies?, full stop.
• performance/get-in-literals: Prefer (-> m :k1 :k2 :k3) over (get-in m [:k1 :k2 :k3]).
• performance/get-keyword: Prefer (:k m) over (get m :k).
• style/redundant-regex-constructor: Prefer #"abc" over (re-pattern #"abc").

Added

• Implemented faster/more efficient versions of Clojure standard library functions:
  • ->list: concrete list building instead of apply . Useful anywhere a lazy-seq might be returned otherwise. seq/vec input: 40/43 us -> 28/15 us
  • mapv*: mapv but short-circuits empty input and uses object-array. Still unsure of this one. 36 us -> 36 us
  • run!*: run! but short-circuits empty input and uses .iterator to perform the side-effects. Does not support reduced. 7 us -> 950 ns
  • pmap*: Avoids lazy-seq overhead and relies on Java's built-in Executors. 3.34 s -> 202 ms
  • walk* and postwalk*: Primarily useful in replace, but may prove useful otherwise. Only supports simple-type defined types. 72 us -> 25 us
• splint.config/read-project-file returns a map of :clojure-version and :paths, taken from the project file (deps.edn or project.clj) in the current directory. If no file is found, :paths is nil and :clojure-version is pulled from *clojure-version*.
• :min-clojure-version in defrule, allowing for rules to specify the minimum version of clojure they require. Rules that are below the supported version are disabled at preparation time and can't be enabled during a run. Acceptable shape is a map of at least one of :major, :minor, and :incremental.
  • Include this in rule documentation.
• test-helpers/with-temp-file and test-helpers/print-to-file! to test file contents.

Changed

• Move spat.parser/parse-string and spat.parser/parse-string-all into the test-helper namespace, and replace with parse-file which accepts the file-obj map.
• Parse data reader/tagged literals as maps instead of lists, and put the extension (dialect) into the symbol's metadata.
• Defer building cli summary until needed.
• Use new splint.config/slurp-edn to read config files, parsed with edamame.
• Changed :spat/lit metadata to :splint/lit. :spat/lit still works for the time being, but no promises.
• splint.printer/print-results now accepts only the results object, which should additionally have :checked-files and :total-time.
• Output formats simple, full, and clj-kondo now print the number of files checked as well: "Linting took 1ms, checked 3 files, 3 style warnings"
• Moved splint.replace/revert-splint-reader-macros into splint.printer where it belongs.
• Rely on undefined behavior in symbol to correctly print unprintable special characters by converting sexprs to strings and then converting those to symbols.
• Move simple-type and drop-quote to splint.utils.

v1.9.0

New Rules

• style/prefer-clj-string: Prefer clojure.string functions over raw interop. Defaults to true.

Added

• --[no]summary cli flag to print or not print the summary line.

Changed

• :filename in Diagnostic is now a java.io.File object, not a string. This is propogated through everything. I suspect no one is using these so I think I could change the Diagnostic as well, but maybe I'll wait a min.
• make-edamame-opts now accepts both features and ns-state, and parse-string and parse-string-all take in features instead of ns-state.
• The runner tracks the filetype of each file and runs over cljc files twice, both clj and cljs, with their respective sides of the reader conditionals applied.
• Diagnostics are deduped before printing.
• lint/warn-on-reflection only runs in clj files.
• Remove farolero. Didn't provide any benefits over judicious try/catch use. :(
• Extend the matcher-combinators.core/Matcher protocol to java.io.File, making match? work nicely with both strings and file objects.
• Performance improvements by converting rules-by-type from a map of simple-type -> map of rule name -> rule to simple-type -> vec of rule.

Fixed

• Correctly print special characters/clojure.core vars (@, not splint/deref, etc).

v1.8.0

New rules

• lint/warn-on-reflection: Require that (set! *warn-on-reflection* true) is called after the ns declaration at the start of every file. Defaults to false.

Breaking

• Deprecate --config. Add --print-config. No timeline for removal of --config (maybe never?).

Added

• edn / edn-pretty output: Print diagnostics as edn using clojure.core/prn and clojure.pprint/pprint.
• Continue to process files after running into errors during rules checking.

Changed

• Dependencies are updated to latest.
• json and json-pretty keys are now sorted.
• Small performance improvements to patterns.

v1.7.0

New Rules

• metrics/parameter-count: Function parameter vectors shouldn't have more than 4 positional parameters. Has :positional and :include-rest styles (only positional or include & args rest params too?), and :count configurable value to set maximum number of parameters allowed.

Added

• Add Metrics rules to documentation.
• Add -s / --silent command line flag to print literally nothing when running Splint.
• json output: Print diagnostics as json using clojure.data.json.
• json-pretty output: Same as json but prettified with pprint.
• Track processed files in :checked-files.
• Add initial corpus files to handle large-scale tests.

Changed

• Move Splint-specific dev code to proper namespaces in dev/.
• Extract splint.runner/run-impl to decomplect processing cli options and returning a status code from performing the actual config loading and rule building and running.
• Rewrite test helper check-all to properly call the existing architecture instead of mock it, to accurately test the run-impl flow.
• Use farolero to handle splint.runner errors. Shows no signs of slowing down the app, so will investigate other areas for usage as well.

Fixed

• Only attach parsed defn metadata when fn name exactly matches defn or defn- and second form is a symbol.
• --no-parallel was producing a lazy seq, now consumes to actually check all files. Oops lol.
• Map over top-level forms with nil parent form instead of treating the whole file as a top-level vector of forms. Fixes naming/lisp-case.
• Add pre- and post- attr-maps to defn metadata when parsing defn forms.
• Added license headers where necessary.

v1.6.1

Fixed

• Re-fix deploy script.

v1.6.0

Added

• Multiple self tests for consistency.
• New test runner based on Cognitect test-runner to print better summary and skip printing namespaces.

Changed

• Cleaned up deploy recipe.
• Wrote short descriptions for all empty config.edn rule descriptions.
• Removed tools.cli defaults for --parallel and --output, now those are added later (see #5).

Fixed

• Correctly merge cli and local options (#5).
• Edge cases for lint/if-not-do, style/when-not-do.

v1.5.0

New Rules

• metrics/fn-length: Function bodies shouldn't be longer than 10 lines. Has :body and :defn styles, and :length configurable value to set maximum length.

Added

• Add test-helpers/expect-match to assert on submatches, transition all existing check-X functions to use it instead.
• Track end position of diagnostics.
• Attach location metadata to function "arities" when a defn arg+body isn't wrapped in a list.

Changed

• Parse defn forms in postprocessing and attach as metadata instead of parsing in individual rules.

Fixed

• Fix style/multiple-arity-order with :arglists metadata.
• Fix binding pattern when binding is falsey.
• Skip #(.someMethod %) in lint/fn-wrapper.
• Skip and and or in style/prefer-condp.

v1.4.1

Fixed

• Fix io/resource issue.
• Remove .class files from jar.

v1.4.0

Added

• -v and --version cli flags to print the current version.
• --config TYPE cli flag to print the diff, local, or full configuration.

Fixed

• Fix "Don't know how to create ISeq from: clojure.lang.Symbol" error in splint.rules.helpers.parse-defn when trying to parse ill-formed function definitions.
• "Fix" error messages. Honestly, I'm not great at these so I'm not entirely sure how to best display this stuff.
• Skip #(do [%1 %2]) in style/useless-do, add docstring note about it.

v1.3.2

Fixed

• Babashka compatibility
• Set up Github CI

v1.3.1

Fixed

• Links in docs for style guide.

v1.3.0

New Rules

• naming/single-segment-namespace: Prefer (ns foo.bar) to (ns foo).
• lint/prefer-require-over-use: Prefer (:require [clojure.string ...]) to (:use clojure.string). Accepts different styles in the replacement form: :as, :refer [...] and :refer :all.
• naming/conventional-aliases: Prefer idiomatic aliases for core libraries ([clojure.string :as str] to [clojure.string :as string]).
• naming/lisp-case: Prefer kebab-case over other cases for top-level definitions. Relies on camel-snake-kebab.
• style/multiple-arity-order: Function definitions should have multiple arities sorted fewest arguments to most: (defn foo ([a] 1) ([a b] 2) ([a b & more] 3))

v1.2.4

Fixed

• Parsing bug in lint/fn-wrapper introduced in v1.2.3.

v1.2.3

Added

• *warn-on-reflection* to all rules and rule template.
• Use :spat/import-ns metadata as way to track when a symbol has been imported.

Changed

• Various performance enhancements:
  • Use protocols in noahtheduke.spat.pattern/simple-type for performance.
  • Use volatile instead of atom for bindings in noahtheduke.spat.pattern.
  • Switch keep to reduce to avoid seq and laziness manipulation.
  • Use some-> where appropriate for short-circuiting.

Fixed

• Fix #2, false positive on interop fn-wrappers.
• Lots of small namespace parsing fixes.

v1.2.2

Fixed

• Differentiate between &&. rest args and parsed lists in :on-match handlers by attaching :noahtheduke.spat.pattern/rest metadata to bound rest args.
• Bump edamame to v1.3.21 to handle #:: {:a 1} auto-resolved namespaced maps with spaces between the colons and the map literal.
• Use correct url in install docs. (Thanks @dpassen)

v1.2.1

Added

• lint/thread-macro-one-arg supports :inline and :avoid-collections styles.
• :updated field in configuration edn, show in rule docs.
• :guide-ref for style/prefer-clj-math.
• Interpose <hr> between each rule's docs.

Changed

• Clarify docstring for lint/dorun-map.

Fixed

• Left align contents of tables in rule docs.
• Correctly render bare links in rule docs.
• Correctly export clojars info in deploy justfile recipe.

v1.2.0

Added

• markdown output: Same text as full but with a fancy horizontal bar, header, and code blocks.
• :chosen-style allows for rules to have configuration and different "styles". The first supported is lint/not-empty? showing either seq or not-empty.

Changed

• ctx is no longer an atom, but a plain map. The :diagnostics entry is now the atom.
• splint.runner/check-form returns the entire updated ctx object instead of just the diagnostics. (I'm not entirely sure that's reasonable, but it's easily changed.)
• Move a lot of rules from lint to style genre:
  • apply-str
  • apply-str-interpose
  • apply-str-reverse
  • assoc-assoc
  • conj-vector
  • eq-false
  • eq-nil
  • eq-true
  • eq-zero
  • filter-complement
  • filter-vec-filterv
  • first-first
  • first-next
  • let-do
  • mapcat-apply-apply
  • mapcat-concat-map
  • minus-one
  • minus-zero
  • multiply-by-one
  • multiply-by-zero
  • neg-checks
  • nested-addition
  • nested-multiply
  • next-first
  • next-next
  • not-eq
  • not-nil
  • not-some-pred
  • plus-one
  • plus-zero
  • pos-checks
  • tostring
  • update-in-assoc
  • useless-do
  • when-do
  • when-not-call
  • when-not-do
  • when-not-empty
  • when-not-not

Breaking

• Add ctx as first argument to :on-match functions to pass in config to rules. Update functions in splint.runner as necessary.

v1.1.1

Changed

• Update Rule Documentation.
• Include new documentation in cljdoc.edn

v1.1.0

Added

• Write documentation for rules and patterns.
• Write docstrings for a bunch of noahtheduke.spat.pattern functions.
• Include outside links in config in rules docs.
• Check :spat/lit metadata to treat special symbols in pattern DSL as their literal values.

Changed

• Attempt to resolve predicates in calling namespace first, then in clojure.core, then in noahtheduke.splint.rules.helpers.
• Rename read-dispatch type from :var to :binding.

v1.0.1

Fixed

• Run linting over syntax-quoted forms again.

v1.0

New Rules

• style/def-fn: Prefer (let [z f] (defn x [y] (z y))) over (def x (let [z f] (fn [y] (z y))))
• lint/try-splicing: Prefer (try (do ~@body) (finally ...)) over (try ~@body (finally ...)).
• lint/body-unquote-splicing: Prefer (binding [max mymax] (let [res# (do ~@body)] res#)) over (binding [max mymax] ~@body).

Added

• Use markdownlint to pretty up the markdown in the repo. Will do my best to keep up with it.

Changed

• Add --parallel and --no-parallel for running splint in parallel or not. Defaults to true.
• No longer run linting over quoted or syntax-quoted forms.
• Rely on edamame's newly built-in :uneval config option for :splint/disable.
• Move version from build.clj to resources/SPLINT_VERSION.

Fixed

• naming/record-name: Add :message.
• style/prefer-condp: Only runs if given more than 1 predicate branch.
• style/set-literal-as-fn: Allow quoted symbols in sets.

v0.1.119

Actually wrote out something of a changelog.

New Rules

• lint/duplicate-field-name: (defrecord Foo [a b a])
• naming/conversion-functions: Should use x->y instead of x-to-y.
• style/set-literal-as-fn: Should use (case elem (a b) true false) instead of (#{'a 'b} elem)

Added

• The :new-rule task now creates a test stub in the correct test directory.
• #_:splint/disable is treated as metadata on the following form and disables all rules. #_{:splint/disable []} can take genres of rules ([lint]) and/or specific rules ([lint/loop-do]) and disables those rules. See below (Thoughts and Followup) for discussion and Configuration for more details.

Changed

• defrule now requires the provided rule-name to be fully qualified, and doesn't perform any *ns* magic to derive the genre.
• Add support for specifying :init-type in defrule to handle symbol matching.
• All of the :dispatch reader macros provided by Edamame now wrap their sexps in the appropriate (splint/X sexp) form, to distinguish them from the symbol forms. Aka #(inc %) is now rendered as (splint/fn [%1] (inc %1)), vs the original (fn* ...), or #'x is now (splint/var x) vs (var x). This allows for writing rules targeting the literal form instead of the symbol form, and requires that rule patterns rely on functions in noahtheduke.splint.rules.helpers to cover these alternates.
• Split all rules tests into their own matching namespaces.
• Add noahtheduke.splint.rules.helpers as an autoresolving namespace so rules can use predicates defined within it without importing or qualifying.
• Renamed errors from violation to diagnostic.
• Merge rules configs into rules maps at load-time.

Fixed

• lint/duplicate-field-name wasn't checking that ?fields was a vector before calling count on it.

Thoughts

I want another parser because I want access to comments. Without comments, I can't parse magic comments, meaning I can't enable or disable rules inline, only globally. That's annoying and not ideal. However, every solution I've dreamed up has some deep issue.

• Edamame is our current parser and it's extremely fast (40ms to parse clojure/core.clj) but it drops comments. I've forked it to try to add them, but that would mean handling them in every other part of the parser, such as syntax-quote and maps and sets, making dealing with those objects really hard. :sob:

• Rewrite-clj only exposes comments in the zip api, meaning I have to operate on the zipper objects with zipper functions (horrible and slow). It's nice to rely on Clojure built-ins instead of (loop [zloc zloc] (z/next* ...)) nonsense.

• clj-kondo is faster than rewrite-clj and has a nicer api, but the resulting tree isn't as easy to work with as Edamame and it's slower. Originally built Spat in it and found it to be annoying to use.

• parcera looked promising, but the pre-processing in parcera/ast is slow and operating on the Java directly is deeply cumbersome. The included grammar also makes some odd choices and I don't know ANTLR4 well enough to know how to fix them (such as including the : in keyword strings). Additionally, if I were to switch, I would have to update/touch every existing rule.

Followup

After tinkering with Edamame for a bit, I've found a solution that requires no changes to edamame to support: #_:splint/disable. This style of directive applies metadata to the following form: #_{:splint/disable [lint/plus-one]} (+ 1 x). Edamame normally discard #_/discarded forms, so on Borkdude's recommendation, I use str/replace to convert it at parse-time to metadata. This uses an existing convention and handles the issue of disabling multiple items or disabling for only a certain portion of the file.

v0.1.85

Update readme with some better writing.

New Rules

• dev/sorted-rules-require for internal use only

Changed

• Annotate all rules with :no-doc.
• Rename lint/cond-else to style/cond-else.
• Cleaned up readme.

v0.1.69

Renamed to Splint! Things are really coming together now.

New Rules

• lint/assoc-in-one-arg
• lint/update-in-one-arg
• naming/predicate
• naming/record-name
• style/let-if
• style/let-when
• style/new-object
• style/prefer-boolean
• style/prefer-condp
• style/redundant-let
• style/single-key-in

Added

• Basic CLI.
• Basic config file and config management.
• cljdoc support.
• -M:gen-docs for rule documentation generation and formatting.
• -M:new-rule task to generate a new rule file from a template.
• -M:deploy task to push to clojars.

Changed

• Split main file into multiple files: core functionality to namespaces, each rule to a separate file.
• Rename lint/with-meta-vary-meta to style/prefer-vary-meta.
• Rename lint/thread-macro-no-arg to lint/redundant-call.

v0.1 - 2023-02-16

Initial release of spat, announcement on Clojurian Slack and bbin installation set up. Contains working pattern matching system, a bunch of rules, and a simple runner.