{"entries":[{"lastCommit":{"author":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2023-04-13T13:23:57.000000+01:00"},"committer":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2023-04-13T14:04:51.000000Z"},"hash":"bad36b2d8a19dd566a27e304407038abbe617968","message":"all: move pkg/gen and internal/cmd/qgo to build-ignored files\n\nThese two are code generators which we need to be able to run.\nOne advantage of having them in their own package directories\nused to be that gopls did not support build tags well at all,\nso editing a file with `//go:build ignore` would show many errors.\n\nIn October 2022, gopls added built-in support for `//go:build ignore`,\nsince it's a very common pattern in cases like this.\nWe can now have these generators as standalone files.\nNot only does that reduce the need for more directories,\nbut it also means `go install ./...` no longer installs the binaries\nconfusingly named `gen` and `qgo` in `go env GOBIN`.\n\nWhile here, update the \"generated by\" comments in pkg source files,\nas these files are no longer automatically generated since\nhttps://cuelang.org/cl/536072, and the comment might confuse someone.\nThe `go run` example is also updated to use the standalone file.\n\nFixes #1869.\n\nSigned-off-by: Daniel Martí \nChange-Id: I3976a613cc501ddaa47ad23afe0d5fd6df536267\nReviewed-on: https://review.gerrithub.io/c/cue-lang/cue/+/552599\nReviewed-by: Marcel van Lohuizen \nUnity-Result: CUEcueckoo \nTryBot-Result: CUEcueckoo \n","tree":"313e62afd9549df9b6231d2dcac34725b3ef8a1d"},"mode":"0100644","name":"big.go","size":1850},{"lastCommit":{"author":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-01-23T19:22:53.000000Z"},"committer":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-01-26T09:06:58.000000Z"},"hash":"9adb47a031652f355985addc15d959f0b22d027c","message":"pkg/math/bits: avoid an extra allocation per call\n\nThese functions implement builtins. In the process of converting their\narguments from CUE values to Go values, we already allocate bit.Ints;\nhence, it's fine to reuse the first argument to store the result,\nbecause we don't need to keep the first argument around.\n\nIf keeping the original value is needed by a caller in the future,\nit's always possible to make a copy of it before the call.\n\n │ old │ new │\n │ B/op │ B/op vs base │\n VetInventory 4.601Gi ± ∞ ¹ 4.593Gi ± ∞ ¹ -0.18% (p=1.000 n=1)\n\n │ old │ new │\n │ allocs/op │ allocs/op vs base │\n VetInventory 48.33M ± ∞ ¹ 47.92M ± ∞ ¹ -0.85% (p=1.000 n=1)\n\nSigned-off-by: Daniel Martí \nChange-Id: I6cf680cfae31732985900e25abd05c778ed01f34\nReviewed-on: https://review.gerrithub.io/c/cue-lang/cue/+/1229640\nUnity-Result: CUE porcuepine \nTryBot-Result: CUEcueckoo \nReviewed-by: Matthew Sackman \n","tree":"17494d98562bc75764b22d035d1468e2acf23047"},"mode":"0040000","name":"bits","size":0},{"lastCommit":{"author":{"email":"noam.tzvi.dolovich@gmail.com","name":"Noam Dolovich","when":"2024-04-23T15:48:57.000000+03:00"},"committer":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2024-04-25T07:45:22.000000Z"},"hash":"c8cc46213c88248953d627dba6d89f6959b97cf6","message":"pkg/math: fix MultipleOf algorithm\n\nmath.MultipleOf currently has many false negatives.\n\nThe current algorithm functions as follows.\nGiven math.MultipleOf(x, y):\n\n - use decimal precision of 1\n - divide x by y\n - check if the resulting apd.Condition is inexact, indicating digits\n were discarded\n\nThis is incorrect, because any time the result of the quotient\nrequires more than the provided precision (i.e. 1), we will assume the\nresult is false.\n\nFor example, math.MultipleOf(22, 2) is false because the result is 11,\nrequiring a precision of 2.\n\nTo fix this, the algorithm is adjusted to:\n\n - divide x by y\n - return whether the result is an integer, by ensuring the\n fractional part is exactly 0\n\nFixes #2926.\n\nSigned-off-by: Noam Dolovich \nChange-Id: Ie978ba0880f7260753eb5288b7c0b14de1c9c73c\nReviewed-on: https://review.gerrithub.io/c/cue-lang/cue/+/1193690\nTryBot-Result: CUEcueckoo \nReviewed-by: Daniel Martí \nUnity-Result: CUE porcuepine \n","tree":"edcce57fa2debc2b5c9f9903f2781387d427f7d1"},"mode":"0100644","name":"manual.go","size":3660},{"lastCommit":{"author":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2023-08-06T00:41:25.000000+02:00"},"committer":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2023-09-04T10:27:13.000000Z"},"hash":"fe2f216d70841f3348917c42f5cb4dac4709cb70","message":"pkg/math: reuse Ln2 and Ln10 in Log2E and Log10E\n\nUpstream Go's math package has always used these simple divisions,\nand now that pkg/gen.go can resolve complex expressions via go/types,\nwe can use exactly the same expression in our package.\n\nThis change is a no-op, which is guaranteed by Example_constants.\nThe new code is a lot less verbose and easier to understand, too.\n\nSigned-off-by: Daniel Martí \nChange-Id: I93dbeb26d3b496c724d53c95afad01f8583e56e0\nReviewed-on: https://review.gerrithub.io/c/cue-lang/cue/+/557323\nTryBot-Result: CUEcueckoo \nReviewed-by: Roger Peppe \n","tree":"9c7cb8cbfe369736f363dad34d110c09660863a5"},"mode":"0100644","name":"math.go","size":13374},{"lastCommit":{"author":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2023-08-06T00:35:14.000000+02:00"},"committer":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2023-08-17T08:20:39.000000Z"},"hash":"a1038f7c2f79327da4f0b57f24d87fde2007560d","message":"pkg/math: fix Log2E and Log10E constant expressions\n\nThey were a division of integer constants, which got truncated,\nmeaning that Log2E rounded down to 1 and Log10E to 0.\n\nTo force them to be floating-point constants, like the CUE ones,\nturn the dividend into a floating-point constant.\n\nI found this problem since staticcheck correctly pointed out\na constant integer division which resulted in zero:\n\n math.go:156:11: the integer division '100… / 230…' results in zero (SA4025)\n\nSigned-off-by: Daniel Martí \nChange-Id: I36ee54a7c8aff06368016b855536a9614dbbf229\nReviewed-on: https://review.gerrithub.io/c/cue-lang/cue/+/557322\nTryBot-Result: CUEcueckoo \nUnity-Result: CUE porcuepine \nReviewed-by: Paul Jolly \n","tree":"2acd74d6f100b1aae5ae83c6435dff26d268edfd"},"mode":"0100644","name":"math_test.go","size":1528},{"lastCommit":{"author":{"email":"matthew@cue.works","name":"Matthew Sackman","when":"2024-07-09T09:51:53.000000+01:00"},"committer":{"email":"matthew@cue.works","name":"Matthew Sackman","when":"2024-07-09T11:44:47.000000Z"},"hash":"304530f592605a5432358572c982667381034d22","message":"core/adt: simplify duplicate NumKind and NumberKind consts\n\nkind.go defined both NumKind and NumberKind as IntKind | FloatKind.\n\nPreviously, NumKind was widely used, and NumberKind was not.\nBut NumberKind is:\n (a) the better name;\n (b) is used in cue/types.go, where many of the other consts from\n adt/kind.go are also repeated.\n (c) matches with Number, as used in the CUE language spec.\n\nRename NumKind to NumberKind and remove duplicate definition.\n\nSigned-off-by: Matthew Sackman \nChange-Id: Iefc35d5326879de9c4c5c3e87b50c530e0be99ef\nReviewed-on: https://review.gerrithub.io/c/cue-lang/cue/+/1197440\nReviewed-by: Daniel Martí \nTryBot-Result: CUEcueckoo \nUnity-Result: CUE porcuepine \n","tree":"2d1e5ecd0d81a2fab918623923f3fd74684f4089"},"mode":"0100644","name":"pkg.go","size":13098},{"lastCommit":{"author":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-03-23T11:37:24.000000Z"},"committer":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-04-01T19:17:33.000000Z"},"hash":"9572fbc264980b498cb2e6951d4bb0afdaa0106e","message":"internal/pkg: rely on the adt package to wrap call errors\n\nIt already does this for builtin validator calls,\nlet it do the same for non-validator calls as well.\n\nThis is a net reduction in code size, given that the adt package\nalready has a bit of code to stringify a builtin func name.\nMoreover, a few dozen errors were missing this wrapping entirely,\ngiven that they failed in ways that internal/pkg did not cover.\n\nAs a bonus, we also avoid stuttering in builtin validator calls,\nand we save about 0.4% of allocated objects in the Inventory module.\n\nSigned-off-by: Daniel Martí \nChange-Id: I1a7cc355476833ef70b9a4b11f21c4d639261ae3\nReviewed-on: https://review.gerrithub.io/c/cue-lang/cue/+/1234621\nReviewed-by: Marcel van Lohuizen \nTryBot-Result: CUEcueckoo \n","tree":"9a21c701702941ac06a327e1568f765dabd0c411"},"mode":"0040000","name":"testdata","size":0}],"lastCommit":{"author":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-03-23T11:37:24.000000Z"},"committer":{"email":"mvdan@mvdan.cc","name":"Daniel Martí","when":"2026-04-01T19:17:33.000000Z"},"hash":"9572fbc264980b498cb2e6951d4bb0afdaa0106e","message":"internal/pkg: rely on the adt package to wrap call errors\n\nIt already does this for builtin validator calls,\nlet it do the same for non-validator calls as well.\n\nThis is a net reduction in code size, given that the adt package\nalready has a bit of code to stringify a builtin func name.\nMoreover, a few dozen errors were missing this wrapping entirely,\ngiven that they failed in ways that internal/pkg did not cover.\n\nAs a bonus, we also avoid stuttering in builtin validator calls,\nand we save about 0.4% of allocated objects in the Inventory module.\n\nSigned-off-by: Daniel Martí \nChange-Id: I1a7cc355476833ef70b9a4b11f21c4d639261ae3\nReviewed-on: https://review.gerrithub.io/c/cue-lang/cue/+/1234621\nReviewed-by: Marcel van Lohuizen \nTryBot-Result: CUEcueckoo \n","tree":"9a21c701702941ac06a327e1568f765dabd0c411"}}