weekly.2024.46

releases: weekly.2024.46

weekly.2024.45

releases: weekly.2024.45

weekly.2024.44

releases: weekly.2024.44

weekly.2024.43

releases: weekly.2024.43

weekly.2024.42

releases: weekly.2024.42

weekly.2024.41

releases: weekly.2024.41

0.4.8

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 supp...

weekly.2024.39

releases: weekly.2024.39

weekly.2024.37

releases: weekly.2024.37

weekly.2024.36

releases: weekly.2024.36