.golangci.yml

run:
  # This is needed for precious, which may run multiple instances
  # in parallel
  allow-parallel-runners: true
  go: "1.23"
  tests: true
  timeout: "30m"

linters:
  enable-all: true
  disable:
    # The canonical form is not always the most common form for some headers
    # and there is a small chance that switching existing strings could
    # break something.
    - canonicalheader

    - cyclop
    - dogsled
    - dupl

    # This is probably worthwhile, but there are a number of false positives
    # that would need to be addressed.
    - dupword

    # We don't follow its policy about not defining dynamic errors.
    - err113

    # We often don't initialize all of the struct fields. This is fine
    # generally
    - exhaustruct

    # We tried this linter but most places we do forced type asserts are
    # pretty safe, e.g., an atomic.Value when everything is encapsulated
    # in a small package.
    - forcetypeassert

    - funlen
    - gochecknoglobals
    - gochecknoinits

    # Similar to the exhaustive linter and I don't know that we use these
    # sorts of sum types
    - gochecksumtype

    - gocognit
    - godox

    # This only "caught" one thing, and it seemed like a reasonable use
    # of Han script. Generally, I don't think we want to prevent the use
    # of particular scripts. The time.Local checks might be useful, but
    # this didn't actually catch anything of note there.
    - gosmopolitan

    # Seems too opinionated or at least would require going through all the
    # interfaces we have.
    - inamedparam

    - ireturn

    # We don't use these loggers
    - loggercheck

    # Maintainability Index. Seems like it could be a good idea, but a
    # lot of things fail and we would need to make some decisions about
    # what to allow.
    - maintidx

    # Using a const for every number doesn't necessarily increase code clarity,
    # and it would be a ton of work to move everything to that.
    - mnd

    # Causes panics, e.g., when processing mmerrors
    - musttag

    - nestif

    # Perhaps too opinionated. We do have some legitimate uses of "return nil, nil"
    - nilnil

    - nlreturn

    # We occasionally use named returns for documentation, which is helpful.
    # Named returns are only really a problem when used in conjunction with
    # a bare return statement. I _think_ Revive's bare-return covers that
    # case.
    - nonamedreturns

    - paralleltest
    - prealloc

    # We have very few structs with multiple tags and for the couple we had, this
    # actually made it harder to read.
    - tagalign

    # Deprecated since golangci-lint 1.64.0. The usetesting linter replaces it.
    - tenv

    # We probably _should_ be doing this!
    - thelper

    # We don't follow this. Sometimes we test internal code.
    - testpackage

    - varnamelen

    # This would probably be good, but we would need to configure it.
    - wsl

linters-settings:
  # Please note that we only use depguard for blocking packages and
  # gomodguard for blocking modules.
  depguard:
    rules:
      main:
        deny:
          - pkg: github.com/likexian/gokit/assert
            desc: Use github.com/stretchr/testify/assert

          - pkg: golang.org/x/exp/maps
            desc: Use maps instead.

          - pkg: golang.org/x/exp/slices
            desc: Use slices instead.

          - pkg: golang.org/x/exp/slog
            desc: Use log/slog instead.

          - pkg: google.golang.org/api/compute/v1
            desc: Use cloud.google.com/go/compute/apiv1 instead.

          - pkg: google.golang.org/api/cloudresourcemanager/v1
            desc: Use cloud.google.com/go/resourcemanager/apiv3 instead.

          - pkg: google.golang.org/api/serviceusage/v1
            desc: Use cloud.google.com/go/serviceusage/apiv1 instead.

          - pkg: io/ioutil
            desc: Deprecated. Functions have been moved elsewhere.

          - pkg: k8s.io/utils/strings/slices
            desc: Use slices

          - pkg: math/rand$
            desc: Use math/rand/v2 or crypto/rand as appropriate.

          - pkg: sort
            desc: Use slices instead

  errcheck:
    # Don't allow setting of error to the blank identifier. If there is a legitimate
    # reason, there should be a nolint with an explanation.
    check-blank: true

    exclude-functions:
      # If we are rolling back a transaction, we are often already in an error
      # state.
      - (*database/sql.Tx).Rollback

      # It is reasonable to ignore errors if Cleanup fails in most cases.
      - (*github.com/google/renameio/v2.PendingFile).Cleanup

      # We often do not care if unlocking failed as we are exiting anyway.
      - (*github.com/gofrs/flock.Flock).Unlock

      # We often don't care if removing a file failed (e.g., it doesn't exist)
      - os.Remove
      - os.RemoveAll

  errorlint:
    errorf: true
    asserts: true
    comparison: true

  exhaustive:
    default-signifies-exhaustive: true

  forbidigo:
    # Forbid the following identifiers
    forbid:
      - p: Geoip
        msg: you should use the `GeoIP` qualifier instead
      - p: geoIP
        msg: you should use the `geoip` qualifier instead
      - p: ^hubSpot
        msg: you should use the `hubspot` qualifier instead
      - p: Maxmind
        msg: you should use the `MaxMind` qualifier instead
      - p: ^maxMind
        msg: you should use the `maxmind` qualifier instead
      - p: Minfraud
        msg: you should use the `MinFraud` qualifier instead
      - p: ^minFraud
        msg: you should use the `minfraud` qualifier instead
      - p: "[Uu]ser[iI][dD]"
        msg: you should use the `accountID` or the `AccountID` qualifier instead
      - p: WithEnterpriseURLs
        msg: Use ghe.NewClient instead.
      - p: ^bigquery.NewClient
        msg: you should use mmgcloud.NewBigQueryClient instead.
      - p: ^drive.NewService
        msg: you should use mmgdrive.NewGDrive instead.
      - p: ^filepath.Walk$
        msg: you should use filepath.WalkDir instead as it doesn't call os.Lstat on every entry.
      - p: ^math.Max$
        msg: you should use the max built-in instead.
      - p: ^math.Min$
        msg: you should use the min built-in instead.
      - p: ^mux.Vars$
        msg: use req.PathValue instead.
      - p: ^net.ParseCIDR
        msg: you should use netip.ParsePrefix unless you really need a *net.IPNet
      - p: ^net.ParseIP
        msg: you should use netip.ParseAddr unless you really need a net.IP
      - p: ^pgtype.NewMap
        msg: you should use mmdatabase.NewTypeMap instead
      - p: ^sheets.NewService
        msg: you should use mmgcloud.NewSheetsService instead.
      - p: ^storage.NewClient
        msg: you should use gstorage.NewClient instead. This sets the HTTP client settings that we need for internal use.
      - p: ^os.IsNotExist
        msg: As per their docs, new code should use errors.Is(err, fs.ErrNotExist).
      - p: ^os.IsExist
        msg: As per their docs, new code should use errors.Is(err, fs.ErrExist)
      - p: ^net.LookupIP
        msg: You should use net.Resolver functions instead.
      - p: ^net.LookupCNAME
        msg: You should use net.Resolver functions instead.
      - p: ^net.LookupHost
        msg: You should use net.Resolver functions instead.
      - p: ^net.LookupPort
        msg: You should use net.Resolver functions instead.
      - p: ^net.LookupTXT
        msg: You should use net.Resolver functions instead.
      - p: ^net.LookupAddr
        msg: You should use net.Resolver functions instead.
      - p: ^net.LookupMX
        msg: You should use net.Resolver functions instead.
      - p: ^net.LookupNS
        msg: You should use net.Resolver functions instead.
      - p: ^net.LookupSRV
        msg: You should use net.Resolver functions instead.

  gci:
    sections:
      - standard
      - default
      - prefix(github.com/maxmind/xgb2code)

  gocritic:
    enable-all: true
    disabled-checks:
      # Revive's defer rule already captures this. This caught no extra cases.
      - deferInLoop
      # Given that all of our code runs on Linux and the / separate should
      # work fine, this seems less important.
      - filepathJoin
      # This seems like it could be good, but we would need to update current
      # uses. It supports "--fix", but the fixing is a bit broken.
      - httpNoBody
      # This might be good, but we would have to revisit a lot of code.
      - hugeParam
      # This might be good, but I don't think we want to encourage
      # significant changes to regexes as we port stuff from Perl.
      - regexpSimplify
      # This seems like it might also be good, but a lot of existing code
      # fails.
      - sloppyReassign
      # I am not sure we would want this linter and a lot of existing
      # code fails.
      - unnamedResult
      # Covered by nolintlint
      - whyNoLint

  gofumpt:
    extra-rules: true

  # IMPORTANT: gomodguard blocks _modules_, not arbitrary packages. Be
  # sure to use the module path from the go.mod file for these.
  # See https://github.com/ryancurrah/gomodguard/issues/12
  gomodguard:
    blocked:
      modules:
        - github.com/avct/uasurfer:
            recommendations:
              - github.com/xavivars/uasurfer
            reason: The original avct module appears abandoned.
        - github.com/BurntSushi/toml:
            recommendations:
              - github.com/pelletier/go-toml/v2
            reason: This library panics frequently on invalid input.
        - github.com/pelletier/go-toml:
            recommendations:
              - github.com/pelletier/go-toml/v2
            reason: This is an outdated version.
        - github.com/gofrs/uuid:
            recommendations:
              - github.com/google/uuid
        - github.com/gofrs/uuid/v5:
            recommendations:
              - github.com/google/uuid
        - github.com/satori/go.uuid:
            recommendations:
              - github.com/google/uuid
        - github.com/lib/pq:
            recommendations:
              - github.com/jackc/pgx
            reason: This library is no longer actively maintained.
        - github.com/neilotoole/errgroup:
            recommendations:
              - golang.org/x/sync/errgroup
            reason: This library can lead to subtle deadlocks in certain use cases.
        - github.com/pariz/gountries:
            reason: This library's data is not actively maintained. Use GeoInfo data.
          github.com/pkg/errors:
            reason: pkg/errors is no longer maintained.
        - github.com/RackSec/srslog:
            recommendations:
              - log/syslog
            reason: This library's data is not actively maintained.
        - github.com/ua-parser/uap-go:
            recommendations:
              - github.com/xavivars/uasurfer
            reason: The performance of this library is absolutely abysmal.
        - github.com/ugorji/go:
            recommendations:
              - encoding/json
              - github.com/mailru/easyjson
            reason: This library is poorly maintained. We should default to using encoding/json and use easyjson where performance really matters.
        - github.com/zeebo/assert:
            recommendations:
              - github.com/stretchr/testify/assert
            reason: Use github.com/stretchr/testify/assert
        - gotest.tools/v3:
            recommendations:
              - github.com/stretchr/testify/assert
            reason: Use github.com/stretchr/testify/assert
        - inet.af/netaddr:
            recommendations:
              - net/netip
              - go4.org/netipx
            reason: inet.af/netaddr has been deprecated.
      versions:
        - github.com/jackc/pgconn:
            reason: Use github.com/jackc/pgx/v5
        - github.com/jackc/pgtype:
            reason: Use github.com/jackc/pgx/v5
        - github.com/jackc/pgx:
            version: < 5.0.0
            reason: Use github.com/jackc/pgx/v5

  gosec:
    excludes:
      # G104 - "Audit errors not checked." We use errcheck for this.
      - G104

      # G306 - "Expect WriteFile permissions to be 0600 or less".
      - G306

      # Prohibits defer (*os.File).Close, which we allow when reading from file.
      - G307

      # no longer relevant with 1.22
      - G601

  govet:
    enable-all: true
    # Although it is very useful in particular cases where we are trying to
    # use as little memory as possible, there are even more cases where
    # other organizations may make more sense.
    disable:
      - fieldalignment
    settings:
      shadow:
        strict: true
  lll:
    line-length: 120
    tab-width: 4

  misspell:
    locale: "US"
    extra-words:
      - typo: "marshall"
        correction: "marshal"
      - typo: "marshalling"
        correction: "marshaling"
      - typo: "marshalls"
        correction: "marshals"
      - typo: "unmarshall"
        correction: "unmarshal"
      - typo: "unmarshalling"
        correction: "unmarshaling"
      - typo: "unmarshalls"
        correction: "unmarshals"

  nolintlint:
    allow-unused: false
    allow-no-explanation: ["lll", "misspell"]
    require-explanation: true
    require-specific: true

  revive:
    enable-all-rules: true
    ignore-generated-header: true
    severity: "warning"

    rules:
      # This might be nice but it is so common that it is hard
      # to enable.
      - name: "add-constant"
        disabled: true

      - name: "argument-limit"
        disabled: true

      - name: "cognitive-complexity"
        disabled: true

      - name: "comment-spacings"
        arguments: ["easyjson", "nolint"]
        disabled: false

      # Probably a good rule, but we have a lot of names that
      # only have case differences.
      - name: "confusing-naming"
        disabled: true

      - name: "cyclomatic"
        disabled: true

      # Although being consistent might be nice, I don't know that it
      # is worth the effort enabling this rule. It doesn't have an
      # autofix option.
      - name: "enforce-repeated-arg-type-style"
        arguments: ["short"]
        disabled: true

      - name: "enforce-map-style"
        arguments: ["literal"]
        disabled: false

      # We have very few of these as we force nil slices in most places,
      # but there are a couple of cases.
      - name: "enforce-slice-style"
        arguments: ["literal"]
        disabled: false

      - name: "file-header"
        disabled: true

      # We have a lot of flag parameters. This linter probably makes
      # a good point, but we would need some cleanup or a lot of nolints.
      - name: "flag-parameter"
        disabled: true

      - name: "function-length"
        disabled: true

      - name: "function-result-limit"
        disabled: true

      - name: "line-length-limit"
        disabled: true

      - name: "max-public-structs"
        disabled: true

      # We frequently use nested structs, particularly in tests.
      - name: "nested-structs"
        disabled: true

      # This doesn't make sense with 1.22 loop var changes.
      - name: "range-val-address"
        disabled: true

      # This flags things that do not seem like a problem, e.g. "sixHours".
      - name: "time-naming"
        disabled: true

      # This causes a ton of failures. Many are fairly safe. It might be nice to
      # enable, but probably not worth the effort.
      - name: "unchecked-type-assertion"
        disabled: true

      # This seems to give many false positives.
      - name: "unconditional-recursion"
        disabled: true

      # This is covered elsewhere and we want to ignore some
      # functions such as fmt.Fprintf.
      - name: "unhandled-error"
        disabled: true

      # We generally have unused receivers in tests for meeting the
      # requirements of an interface.
      - name: "unused-receiver"
        disabled: true

  tagliatelle:
    case:
      rules:
        avro: "snake"
        bson: "snake"
        env: "upperSnake"
        envconfig: "upperSnake"
        json: "snake"
        mapstructure: "snake"
        xml: "snake"
        yaml: "snake"

  unparam:
    check-exported: true

  wrapcheck:
    ignoreSigs:
      - ".Errorf("
      - "errgroup.NewMultiError("
      - "errors.Join("
      - "errors.New("
      - ".Wait("
      - ".WithStack("
      - ".Wrap("
      - ".Wrapf("
      - "v5.Retry[T any]("

issues:
  exclude-use-default: false

  exclude-dirs:
    - "geoip-build/mmcsv"

  exclude-files:
    - "_easyjson\\.go$"
    - "_easyjson_test\\.go$"
    - "_xgb2code\\.go$"
    - "_json2vector\\.go$"

  exclude-rules:
    - linters:
        - "bodyclose"
      # This rule doesn't really make sense for tests where we don't have an open
      # connection and we might be passing around the response for other reasons.
      path: "_test.go"

    - linters:
        - "errcheck"
      # There are many cases where we want to just close resources and ignore the
      # error (e.g., for defer f.Close on a read). errcheck removed its built-in
      # wildcard ignore. I tried listing all of the cases, but it was too many
      # and some were very specific.
      source: "\\.Close"

    - linters:
        - "forbidigo"
      # This refers to a minFraud field, not the MaxMind Account ID
      source: "AccountUserID|Account\\.UserID"

    # we include both a source and text exclusion as the source exclusion
    # misses matches where forbidigo reports the error on the first line
    # of a chunk of a function call even though the use is on a later line.
    - linters:
        - "forbidigo"
      text: "AccountUserID|Account\\.UserID"

    - linters:
        - "gocritic"
      # For some reason the imports stuff in ruleguard doesn't work in golangci-lint.
      # Perhaps it has an outdated version or something
      path: "_test.go"
      text: "ruleguard: Prefer the alternative Context method instead"

    - linters:
        - "gocritic"
      # The nolintlint linter behaves oddly with ruleguard rules
      source: "// *no-ruleguard"

    - linters:
        - "nolintlint"
      # The contextcheck linter also uses "nolint" in a slightly different way,
      # leading to falso positives from nolintlint.
      source: "//nolint:contextcheck //.*"

    - linters:
        - "govet"
      # These are usually fine to shadow and not allowing shadowing for them can
      # make the code unnecessarily verbose.
      text: 'shadow: declaration of "(ctx|err|ok)" shadows declaration'

    - linters:
        - "contextcheck"
        # With recent changes to the linter, there were a lot of failures in
        # the tests and it wasn't clear to me that fixing them would actually
        # improve the readability.
        - "goconst"
        - "nilerr"
        - "wrapcheck"
      path: "_test.go"

    - linters:
        - "stylecheck"
      # ST1016 - methods on the same type should have the same receiver name.
      #    easyjson doesn't interact well with this.
      text: "ST1016"

    - linters:
        - "wrapcheck"
      text: "github.com/maxmind/xgb2code"

    - linters:
        - "wrapcheck"
      path: "_easyjson.go"

    - linters:
        - "gocritic"
      source: "Chmod|WriteFile"
      text: "octalLiteral"