vlang/v 0.4.8

on GitHub

V 0.4.8

28 Sep 2024

Improvements in the language

• A new implements keyword for explicit interface implementation
• Allow multi return as fn argument (#21991)
• Define a default sumtype value (based on the first variant type) (#22039)
• Remove the obsolete .code and .msg fields of IError (#22066)
• Fix generic lambda type binding and resolution (#22083)
• Comptime support for traversing the method parameters with $for param in method.params { (#22229)
• Show missing variants in the sum type error
• A much better and detailed unmatched fn arg error
• Add support for @BUILD_DATE, @BUILD_TIME and @BUILD_TIMESTAMP, all using v.util.get_build_time(), and overridable through SOURCE_DATE_EPOCH (#22213)

Breaking changes

• Deprecate x.vweb and vweb in favor of veb, a faster, easier, and more stable framework.

Checker improvements/fixes

• Disallow static fn call when receiver type is unknown (#21970)
• Fix sumtype checking for voidptr variant (#21955)
• Check comptime veb.html('index.html') (#21961)
• Check if a parent generic struct has concrete types or not (#21962)
• Add support for static methods in @FN and @METHOD (#21990)
• Add a deprecation warning for const () groups (an error after 2025-01-01) (#22019)
• Improve -d trace_checker and error diagnostic information on compiler panics
• Add error when initializing sumtype with struct as first type (#22067)
• Add infix checks for nil (#22045)
• Fix map generic fn arg passing (#22071)
• Disallow using a preexisting const name in a for loop, as either a key or value ident (#22108)
• Fix generic lambda type binding resolution (fix #22109) (#22115)
• Fix array alias (#22175)
• Restrict multiple union fields initialised at once with a nicer checker error, instead of producing an enigmatic error at cgen time (#22196)
• Fix compilation of vlib/v/slow_tests/assembly/asm_test.amd64.v (regression after dfc0c91)
• Add missing check for ref passing to non-ref (#22194)
• Check struct implements non interface type (fix #22200) (#22218)
• Suggest using the @[_allow_multiple_values] attribute, when declaring enums that have duplicate values (#22224)
• Check for duplicate interface names in the implements parts of struct declarations (#22230)
• Fix missing struct field type checking for type mismatch (ref vs non-ref in mt sync.Mutex = sync.new_mutex()) (fix #18088) (#21949)
• Fix fntype var marked as auto heap (#22290)
• Check array.delete() argument mismatch (#22307)
• Add missing check for duplicated items on in expr (fix #22305) (#22308)
• Disallow infix expr on left side of assign (#22322)
• Fix array fixed return type for interface methods (#22320)
• Check arguments mismatch of array.sorted_with_compare() (fix #22327) (#22328)
• Add an error for returning an any value in pure V code (prevents invalid cgen) (fix #12623) (#22334)
• Cleanup the checking of array method calls (#22338)
• Fix voidptr type checking (#21923)

Parser improvements

• Fix lots of parser panics, discovered through fuzzing with radamsa
• Improve the error for keyword lock, used as a variable name (#21937)
• Improve the error message position for invalid array attr keys (#21944)
• Fix const field str() (#21998)
• Update @include in templates, to work with relative paths & prevent recursive calls (#21943)
• Check fn call args without comma between them (related #22021) (#22075)
• parser,scanner,ast: make the scanner and parser more robust, by implementing more limits (preventing panics, discovered by fuzzing)
• Protect against too deep recursion in Expr.pos() calls
• Check too many layers embedded generic type (fix #22089) (#22091)
• Cache ident lookups for consts in ast Expr str (#22101)
• Improve Type and TypeFlag related operations (#22107)
• Fix parsing map value inside or expr (fix #12164) (#22180)
• Fix const field str() (#22192)
• Fix .${var} used in a template, compiled by $tmpl() (fix #22231) (#22270)
• Check enum method duplicated (fix #20924) (#22294)

Compiler internals

• scanner: guard against scanner panic, discovered by fuzzing in PR#22016
• v.builder: show the thirdparty object compilation commands too, when using -showcc (when the cache is empty)
• builder: allow for v -dump-defines - -check cmd/v, which is faster, because it can skip code generation
• Reduce allocations for the most common cases (#22142)
• transformer: add support for instrumenting the V compiler with -d trace_transformer

Standard library

• encoding.base58: fix notice for slice creation (#21935)
• gg: reset ctx.mouse_d? and ctx.scroll_? at the end of each frame (fix #21945) (#21946)
• builtin: v_segmentation_fault_handler signal_number i32
• builtin: fix 'aaaa'.split('aa') (fix #21936) (#21951)
• builtin: panic on trying to grow arrays with capacity bigger than 2^31, instead of overflowing a.cap (partial fix for #21918) (#21947)
• gg: add a note that Context.new_streaming_image has to be called after Sokol's setup
• gg: add more documentation comments for gg.Config (the parameters of gg.start and gg.new_context)
• regex: fix regex.split() (fix #16876) (#21953)
• json: increase test cases before enabling sumtype decode in all json libraries (#21958)
• gg: change the type of gg.DrawImageConfig.rotate from int to f32
• gg: deprecate gg.DrawImageConfig.rotate, in favor of gg.DrawImageConfig.rotation, improve the documentation comments (#21963)
• x.crypto.chacha20: make Cipher struct public (fix #21967) (#21968)
• tmpl: fix an extra newline in @for; builtin: some i64 fixes
• gg: add an optional size: parameter to the .draw_pixels and .draw_pixel methods (defaults to 1.0)
• sokol: update to match upstream at c0e0563 (#21971)
• Add support for Any in decode_struct, encode_struct and to_any (#21972)
• crypto.cipher: make Stream.xor_key_stream implementers require a mutable receiver (#21974)
• sokol.audio: fix ./v -cc clang-18 -gc none simple_bytebeat.v (the audio depends on threads)
• time: d,c,dd,ddd,dddd pattern support for parse_format() (#22003)
• flag: add optional value description to string parameters (#22024)
• flag: add custom value descriptions for bool, int, and float flags too (#22032)
• flag: fix assigning to @[tail] field when no fields has been matched yet in flag.parse[T]() (#22043)
• crypto: add a crypto.pbkdf2 module (#22047)
• hash: add more methods to the hash.Hash interface, to match the ones in Go (#22001)
• arrays: simplify arrays.sum and arrays.reduce (#22076)
• x.json2: support @[skip] as well (#22077)
• builtin,thirdparty: fix compilation of libgc with -cc msvc -gc boehm (thanks to @Ekopalypse)
• stbi: change Image.data from voidptr to &u8, to reduce casts (#21977)
• time: update parse_format comment description in parse.c.v (#22104)
• vlib: add an arrays.parallel module, containing parallel.run/3 and parallel.amap/3 implementations (#22090)
• builtin: support -d builtin_print_use_fprintf, make the C fn declarations stricter (#22137)
• builtin: fix map.clear() not resetting map's metas and keys blocks (fix #22139) (#22140)
• builtin: fix incomplete m.clear(), allowing the map to have a duplicated entry for its first key (fix #22143) (#22144)
• builtin: fix m.clear() having different observable behavior to m = {}, after multiple iterations of setting keys and clearing (fix #22145) (#22146)
• builtin: fix bug in .clear() caused by sizeof(u32) being 4, not 2 (fix #22148)
• flag: add support for parsing flag.FlagParser style flags in to_struct[T] (#22152)
• flag: fix parse_bool_value() (#22160)
• flag: correct bool logic, add test (#22162)
• flag: fix parsing flag.FlagParser style short flags in to_struct[T] (#22172)
• gg: change the type of PenConfig.thickness to f32
• builtin: remove remaining references to v_calloc in function comments (#22179)
• builtin: remove string interpolation from panic/1, to be able to use tools like cbmc in more cases (#22182)
• flag: add a relaxed parsing mode, that turn flag match errors into no_match entries instead (#22191)
• encoding.binary: add u16/u32/u64 -> []u8 conversion functions (#22193)
• crypto.sha1, crypto.sha256, crypto.sha3, crypto.sha512: improve performance for non prod builds, by tagging the block_generic functions with @[direct_array_access]
• builtin: fix string.trim() (fix #13021) (#22205)
• crypto.bcrypt: reduce runtime cost for running bcrypt_test.v, by reducing the iteration count
• crypto.scrypt: add a new scrypt module to vlib/crypto (#22216)
• sync.stdatomic: add OpenSUSE paths for libatomic
• crypto.scrypt: add missing comment of source for test vector (#22222)
• json: allow passing an anon struct as a decode type (#22228)
• flag: fix parse_bool_value() with different order short args (fix #22176) (#22242)
• builtin: drop C in int.v (#22245)
• strconv: fix format_fl()/format_es() (fix #13210) (#22244)
• json: fix decoding of structs with embeds (#22264)
• crypto.rand: add support for convenient generation of a random big integer in the interval [0, n) (#22266)
• json: fix json encode/decode with embed support (#22277)
• io: add a BufferedWriter and supporting methods (#22265)
• vlib: add a go like x.benchmark module, that estimates automatically how many iterations are needed, to get a statistically significant result (#22215)
• math: document q_rsqrt
• io: make buffered_writer_test.v more robust
• builtin: enable GC lib on rv64 build (#22319)
• json: support null sum types in decode()
• crypto: ecdsa module (on top of openssl)
• bench: crypto/ecdsa.v
• math.big: fix a + b and a - b, when the signs are different, add more test cases (#22330)

Web

• Check for using comptime $veb.html()/$vweb.html(), without importing veb or vweb (#21957)
• net: add net.Dialer and net.Connection interfaces, abstracting the different types of connections, already supported by the V network stack (#21657)
• net.mbedtls: support Server Name Indication (SNI) (#22012)
• veb: extract constants into consts.v (#22132)
• vweb: mark vweb as deprecated in its README, recommending using veb instead (#22131)
• veb: fix vweb_livereload reference to veb_livereload (#22171)
• veb: fix a few minor errors in the README.md (#22177)
• net.mbedtls: store the client ip (for ipv4), shutdown on handshake failure, in .accept() (#22184)
• veb: implicit context
• veb: make implicit context work with custom user types
• net.websocket: use retrying on EINTR in Client.listen() (#22279)
• net: allow ipv6 address with brackets (fix #22313) (#22316)

Database drivers

• db.sqlite: add instructions for installing SQLite's amalgamation or development package, if it is missing

C backend

• Fix struct ref field with no ref structinit (#21932)
• Define a flexible vint_t type (32 bit int on 32 bit systems, 64 bit otherwise)
• Fix generic sumtype with repeated concrete type (#21948)
• Fix array.map with closure var fn (#22002)
• Fix generation of closures from methods on aliases (#22004)
• Reduce indentation level for generated defer statements
• Fix selector with interface var (#22006)
• Fix free method generation for option struct (#22060)
• Fix fixed array with default init (#22059)
• Fix for loop with array fixed returned from fn (#22069)
• Fix free method for option fields (#22070)
• Fix auto free method for option map (fix #22081) (#22094)
• Return early from autofree related functions, when -autofree is not used
• Fix shared object method call (fix #22121) (#22125)
• Fix array fixed initialization from map indexing (fix #22133) (#22149)
• Fix generic options with reserved ident (#22164)
• Fix struct field with default optional value (fix #11119) (#22167)
• Fix array of fns index call with embeded array index (fix #17381) (#22198)
• Fix match with mut cond variable (#22207)
• Fix code generated for indexexpr with complex assigning (#22203)
• Fix interface type generation for generic struct (fix #22186) (#22189)
• Fix wrong type of vint_t and const riscv64 (#22251)
• Fix code for C ident when ptr is expected (#22259)
• Fix C name mangling with inherited closure vars (fix #22262) (#22263)
• Fix codegen for alias to charptr passed as ptr (fix #9679) (#22271)
• Fix codegen for option fntype used in a match (fix #22278) (#22280)
• Fix option cast from fntype (fix #22283, #22284) (#22285)
• Fix codegen for method call on rangeexpr (fix #12610) (#22291)
• Fix operation overload for type aliases of fixed arrays (fix #22297) (#22304)
• Fix codegen for assign from unsafeexpr resulting fixed array (fix #22301) (#22309)
• Fix variadic arg var passed to another call which expects variadic (fix #22315) (#22317)
• Fix aliases of array insert(...)/prepend(...) (fix #22323) (#22324)
• Fix codegen for interface method call which returns a fixed array (fix #22326) (#22331)
• Add asm to c_reserved, fixes compilation of struct Abc { @asm int } (#22340)
• Fix interface method call after smartcast (fix #17056) (#22335)
• Fix codegen for address of range (fix #18528) (#22336)

vfmt

• Allow align threshold to be parametrized in calls to add_new_info (#21942)
• Fix and simplify align of struct fields (#21995)
• Fix alignment of enumeration types (#21999)
• Fix enum fields with one empty line (#22007)
• Fix fmt of enum fields with empty line (#22015)
• Fix alignment of struct init fields (#22025)
• Keep empty newlines in between interface fields/methods (#22040)
• Fix interface fields or methods with empty newlines (#22046)
• Fix enum/struct_decl/struct_init fields with empty newlines (#22051)
• Fix interface fields/methods alignment (#22055)
• Remove the prefixed module name of const names, that are in the same module (related #22183) (#22185)
• Fix import selective with interface implements (fix formatting part of #22200) (#22209)
• Add a test for fn with c binding type args (#22212)
• Fix formating a file in an oscillating manner (fix #22223, fix #22026) (#22232)

Tools

• Implement a -repeats/-R N option to v repeat, to eliminate another need for platform dependent scripting
• ci: add hub_docker_ci.yml, for building docker images (triggered manually for now) (#22302)
• ci: use docker compose instead of docker-compose (see https://github.com/orgs/community/discussions/116610) (#21992)
• vrepl: suppress the welcome message, if VQUIET is set (#21941)
• Make v where ignore .git/ folders (they contain binary files); let -dir . work recursively, so that -dir vlib works
• Sort the match results in the vwhere test, to make it less flaky (#22033)
• Add an amalgamate tool and description of usage (#22034)
• Add a few missing v command entries and their flags in v complete (#22041)
• Colorise the output of cmd/tools/show_ancient_deprecations.v, reduce false positives (#22048)
• docs: clarify the .precision specification section for string interpolation of floats (#22061)
• docs: add a copy code function (top/right copy icon) on doc examples (#22114)
• Allow passing parameters to fuzz_v_parser_with_radamsa.sh, so that it could be run in several processes from the same folder
• Use separate .autofuzz.log files too for the separate invocations of fuzz_v_parser_with_radamsa.sh
• Extract .github/workflows/run_sanitizers.sh to ease local testing with different options
• parser,ast: protect against more overflows/panics, by forcing early returns on deeply nested expressions and scopes (#22098)
• Improve the output of v repeat command, by coloring the faster/slower label
• Add diagnostic in v repeat for invalid combinations of -r, -i and -a flags
• Fix v doc truncating code blocks, that lack a specific language (fix #22017)
• v.util: add get_build_time/0, supporting https://reproducible-builds.org/docs/source-date-epoch/
• Fix v doc not converting <s> in plain code blocks into encoded html entities in its .html output
• ci: run npx prettier --write **.yml; ensure it is run on all .yml files, not just the ones in the .github/workflows/ folder
• docs: add implements keyword for explicit interface implementations (#22214)
• Make fast_job.v more robust (setup a custom PATH) and informative on fast.v failures (compile it with -g)
• Make fast.v and fast_job.v more self sufficient
• ci: add cache and trigger for pushes, in the Hub docker action (#22314)

Operating System support

• docs: streamline the installation instructions and notes for Windows and Ubuntu
• v.builder: fix errors in cstrict mode on OpenBSD with clang (#22154)
• thirdparty: fix compilation of programs using miniz.h on macos
• crypto.rand: fix compilation on macos with -cc tcc -no-retry-compilation -gc none
• thirdparty: fix compilation of thirdparty/mbedtls with tcc on macos m1
• v.pkgconfig: fix parser, when includedir= lines, had trailing spaces (fix -d use_openssl for openssl 3.3.2 installed through brew on macos)
• builtin: fix compilation with tcc on OpenBSD using libgc (#22234)
• docs: add FreeBSD in cross compilation section (#22249)
• v.builder: enable LTO for clang on OpenBSD (#22247)
• thirdparty: fix compilation of programs using miniz.h on OpenBSD (#22254)
• net: fix compilation on windows (use casts to int for net.error_eintr etc)
• net: use explicit int casts for net.error_eintr etc in the unix implementation too for consistency with the windows one

Examples

• Remove drag_n_drop.v from the list of examples, that are checked for visual differences with vgret, since it now uses the default gg font
• docs: add more C interop notes and examples (#21965)
• cleanup obsolete unsafe{} usages in examples/sokol/sounds
• cleanup & fix the sound produced by melody.v
• add a simplified bytebeat player to show how to use sokol.audio, and that does not depend on gg
• make rotating_textured_quad.v compile and run on Android (#21987)
• veb: change example description, to avoid repetitive wording (ease debugging of issue#22017)
• eval: fix ./v interpret examples/hanoi.v
• add examples/ttf_font/draw_static_text.v, to make it easier to test x.ttf with different fonts, and texts
• make draw_static_text.v show the font name too, for easier comparisons
• prevent cliping of long font names in draw_static_text.v
• docs: add an example on how to use Options/Results, when returning multiple values from a function (#22099)
• add examples/gg/draw_unicode_text_with_gg.v, for easy comparison of how different fonts and unicode texts will look, when rendered by gg
• add examples/veb/websocket, to show how to use http connection upgrade to a websocket, from a veb route (#22128)
• migrate vweb examples to veb
• fix type in veb_example.v
• add gc_is_enabled() check to 2048 to prevent crash in Android emulator (#22274)
• make gg/rectangles.v Android friendly (#22275)
• make gg/stars.v run on android (#22276)
• examples,os: add an os.asset module, use it to simplify code in examples/, by removing $if android { checks (#22281)
• add a consistent background to flappylearning, shown when the height of the view is very high (on Android)