with_ascii_lowercased zig builtin

Author: HajagosNorbert

crates/compiler/builtins/bitcode/src/main.zig

@@ -211,6 +211,7 @@ comptime {
     exportStrFn(str.withCapacityC, "with_capacity");
     exportStrFn(str.strAllocationPtr, "allocation_ptr");
     exportStrFn(str.strReleaseExcessCapacity, "release_excess_capacity");
+    exportStrFn(str.strWithAsciiLowercased, "with_ascii_lowercased");
 
     for (INTEGERS) |T| {
         str.exportFromInt(T, ROC_BUILTINS ++ "." ++ STR ++ ".from_int.");

crates/compiler/builtins/bitcode/src/str.zig

@@ -2,6 +2,7 @@ const utils = @import("utils.zig");
 const RocList = @import("list.zig").RocList;
 const UpdateMode = utils.UpdateMode;
 const std = @import("std");
+const ascii = std.ascii;
 const mem = std.mem;
 const unicode = std.unicode;
 const testing = std.testing;
@@ -374,7 +375,12 @@ pub const RocStr = extern struct {
             return utils.REFCOUNT_ONE;
         }
 
-        const ptr: [*]usize = @as([*]usize, @ptrCast(@alignCast(self.bytes)));
+        const data_ptr = if (self.isSeamlessSlice())
+            self.getAllocationPtr()
+        else
+            self.bytes;
+
+        const ptr: [*]usize = @as([*]usize, @ptrCast(@alignCast(data_ptr)));
         return (ptr - 1)[0];
     }
 
@@ -615,16 +621,6 @@ fn initFromSmallStr(slice_bytes: [*]u8, len: usize, _: usize) RocStr {
     return RocStr.init(slice_bytes, len);
 }
 
-// The alloc_ptr must already be shifted to be ready for storing in a seamless slice.
-fn initFromBigStr(slice_bytes: [*]u8, len: usize, alloc_ptr: usize) RocStr {
-    // Here we can make seamless slices instead of copying to a new small str.
-    return RocStr{
-        .bytes = slice_bytes,
-        .length = len | SEAMLESS_SLICE_BIT,
-        .capacity_or_alloc_ptr = alloc_ptr,
-    };
-}
-
 fn strSplitOnHelp(array: [*]RocStr, string: RocStr, delimiter: RocStr) void {
     if (delimiter.len() == 0) {
         string.incref(1);
@@ -1972,6 +1968,66 @@ fn countTrailingWhitespaceBytes(string: RocStr) usize {
     return byte_count;
 }
 
+// Str.with_ascii_lowercased
+pub fn strWithAsciiLowercased(string: RocStr) callconv(.C) RocStr {
+    var new_str = if (string.isUnique())
+        string
+    else blk: {
+        string.decref();
+        break :blk RocStr.fromSlice(string.asSlice());
+    };
+
+    const new_str_bytes = new_str.asU8ptrMut()[0..string.len()];
+    for (new_str_bytes) |*c| {
+        c.* = ascii.toLower(c.*);
+    }
+    return new_str;
+}
+
+test "withAsciiLowercased: small str" {
+    const original = RocStr.fromSlice("cOFFÉ");
+    try expect(original.isSmallStr());
+
+    const expected = RocStr.fromSlice("coffÉ");
+    defer expected.decref();
+
+    const str_result = strWithAsciiLowercased(original);
+    defer str_result.decref();
+
+    try expect(str_result.isSmallStr());
+    try expect(str_result.eq(expected));
+}
+
+test "withAsciiLowercased: non small str" {
+    const original = RocStr.fromSlice("cOFFÉ cOFFÉ cOFFÉ cOFFÉ cOFFÉ cOFFÉ");
+    defer original.decref();
+    try expect(!original.isSmallStr());
+
+    const expected = RocStr.fromSlice("coffÉ coffÉ coffÉ coffÉ coffÉ coffÉ");
+    defer expected.decref();
+
+    const str_result = strWithAsciiLowercased(original);
+
+    try expect(!str_result.isSmallStr());
+    try expect(str_result.eq(expected));
+}
+
+test "withAsciiLowercased: seamless slice" {
+    const l = RocStr.fromSlice("cOFFÉ cOFFÉ cOFFÉ cOFFÉ cOFFÉ cOFFÉ");
+    const original = substringUnsafeC(l, 1, l.len() - 1);
+    defer original.decref();
+
+    try expect(original.isSeamlessSlice());
+
+    const expected = RocStr.fromSlice("offÉ coffÉ coffÉ coffÉ coffÉ coffÉ");
+    defer expected.decref();
+
+    const str_result = strWithAsciiLowercased(original);
+
+    try expect(!str_result.isSmallStr());
+    try expect(str_result.eq(expected));
+}
+
 fn rcNone(_: ?[*]u8) callconv(.C) void {}
 
 fn decStr(ptr: ?[*]u8) callconv(.C) void {

crates/compiler/builtins/roc/Str.roc

@@ -369,6 +369,7 @@ module [
     contains,
     drop_prefix,
     drop_suffix,
+    with_ascii_lowercased,
 ]
 
 import Bool exposing [Bool]
@@ -1092,3 +1093,18 @@ drop_suffix = \haystack, suffix ->
         substring_unsafe(haystack, start, len)
     else
         haystack
+
+## Returns a version of the string with all [ASCII characters](https://en.wikipedia.org/wiki/ASCII) lowercased. Non-ASCII characters are left unmodified. For example:
+##
+## ```roc
+## expect "CAFÉ".with_ascii_lowercased() == "cafÉ"
+## ```
+##
+## This function is useful for things like [command-line options](https://en.wikipedia.org/wiki/Command-line_interface#Command-line_option) and [environment variables](https://en.wikipedia.org/wiki/Environment_variablewhere you ## know in advance that you're dealing with a hardcoded string containing only ASCII characters. It has better performance than lowercasing operations which take Unicode into account.
+##
+## That said, strings received from user input can always contain non-ASCII Unicode characters, and lowercasing [Unicode](https://unicode.org) works differently in different languages. For example, the string `"I"lowercases to `"i"## ` in English and to `"ı"` (a [dotless i](https://en.wikipedia.org/wiki/Dotless_I)) in Turkish. These rules can also change in each [Unicode release](https://www.unicode.org/releases/), so we have separate [`unicode` package]## (https://github.com/roc-lang/unicode) for Unicode capitalization that can be upgraded independently from the language's builtins.
+##
+## To do a case-insensitive comparison of the ASCII characters in a string, use [`caseless_ascii_equals`](#caseless_ascii_equals).
+with_ascii_lowercased: Str -> Str
+
+expect Str.with_ascii_lowercased "cOFFÉ" == "XYZFÉ"

crates/compiler/builtins/src/bitcode.rs

@@ -358,6 +358,7 @@ pub const STR_CLONE_TO: &str = "roc_builtins.str.clone_to";
 pub const STR_WITH_CAPACITY: &str = "roc_builtins.str.with_capacity";
 pub const STR_ALLOCATION_PTR: &str = "roc_builtins.str.allocation_ptr";
 pub const STR_RELEASE_EXCESS_CAPACITY: &str = "roc_builtins.str.release_excess_capacity";
+pub const STR_WITH_ASCII_LOWERCASED: &str = "roc_builtins.str.with_ascii_lowercased";
 
 pub const LIST_MAP: &str = "roc_builtins.list.map";
 pub const LIST_MAP2: &str = "roc_builtins.list.map2";

crates/compiler/can/src/builtins.rs

@@ -130,6 +130,7 @@ map_symbol_to_lowlevel_and_arity! {
     StrToNum; STR_TO_NUM; 1,
     StrWithCapacity; STR_WITH_CAPACITY; 1,
     StrReleaseExcessCapacity; STR_RELEASE_EXCESS_CAPACITY; 1,
+    StrWithAsciiLowercased; STR_WITH_ASCII_LOWERCASED; 1,
 
     ListLenUsize; LIST_LEN_USIZE; 1,
     ListLenU64; LIST_LEN_U64; 1,

crates/compiler/gen_dev/src/lib.rs

@@ -1726,6 +1726,13 @@ trait Backend<'a> {
                 arg_layouts,
                 ret_layout,
             ),
+            LowLevel::StrWithAsciiLowercased => self.build_fn_call(
+                sym,
+                bitcode::STR_WITH_ASCII_LOWERCASED.to_string(),
+                args,
+                arg_layouts,
+                ret_layout,
+            ),
             LowLevel::StrToNum => {
                 let number_layout = match self.interner().get_repr(*ret_layout) {
                     LayoutRepr::Struct(field_layouts) => field_layouts[0], // TODO: why is it sometimes a struct?

crates/compiler/gen_llvm/src/llvm/lowlevel.rs

@@ -593,6 +593,7 @@ pub(crate) fn run_low_level<'a, 'ctx>(
                 bitcode::STR_WITH_CAPACITY,
             )
         }
+
         ListLenU64 => {
             // List.len : List * -> U64
             arguments!(list);
@@ -635,6 +636,17 @@ pub(crate) fn run_low_level<'a, 'ctx>(
                 list_element_layout!(layout_interner, result_layout),
             )
         }
+        StrWithAsciiLowercased => {
+            arguments!(string);
+
+            call_str_bitcode_fn(
+                env,
+                &[string],
+                &[],
+                BitcodeReturns::Str,
+                bitcode::STR_WITH_ASCII_LOWERCASED,
+            )
+        }
         ListConcat => {
             debug_assert_eq!(args.len(), 2);
 

crates/compiler/gen_wasm/src/low_level.rs

@@ -258,6 +258,9 @@ impl<'a> LowLevelCall<'a> {
                 self.load_args_and_call_zig(backend, bitcode::STR_SUBSTRING_UNSAFE)
             }
             StrWithCapacity => self.load_args_and_call_zig(backend, bitcode::STR_WITH_CAPACITY),
+            StrWithAsciiLowercased => {
+                self.load_args_and_call_zig(backend, bitcode::STR_WITH_ASCII_LOWERCASED)
+            }
 
             // List
             ListLenU64 => {

crates/compiler/module/src/low_level.rs

@@ -26,6 +26,7 @@ pub enum LowLevel {
     StrReserve,
     StrWithCapacity,
     StrReleaseExcessCapacity,
+    StrWithAsciiLowercased,
     ListLenUsize,
     ListLenU64,
     ListWithCapacity,
@@ -267,6 +268,7 @@ map_symbol_to_lowlevel! {
     StrToNum <= STR_TO_NUM;
     StrWithCapacity <= STR_WITH_CAPACITY;
     StrReleaseExcessCapacity <= STR_RELEASE_EXCESS_CAPACITY;
+    StrWithAsciiLowercased <= STR_WITH_ASCII_LOWERCASED;
     ListLenU64 <= LIST_LEN_U64;
     ListLenUsize <= LIST_LEN_USIZE;
     ListGetCapacity <= LIST_CAPACITY;

crates/compiler/module/src/symbol.rs

@@ -1418,6 +1418,7 @@ define_builtins! {
         48 STR_RELEASE_EXCESS_CAPACITY: "release_excess_capacity"
         49 STR_DROP_PREFIX: "drop_prefix"
         50 STR_DROP_SUFFIX: "drop_suffix"
+        51 STR_WITH_ASCII_LOWERCASED: "with_ascii_lowercased"
     }
     6 LIST: "List" => {
         0 LIST_LIST: "List" exposed_apply_type=true // the List.List type alias

crates/compiler/mono/src/drop_specialization.rs

@@ -1549,6 +1549,7 @@ fn low_level_no_rc(lowlevel: &LowLevel) -> RC {
         ListPrepend => RC::Rc,
         StrJoinWith => RC::NoRc,
         ListSortWith => RC::Rc,
+        StrWithAsciiLowercased => RC::Rc,
 
         ListAppendUnsafe
         | ListReserve

crates/compiler/mono/src/inc_dec.rs

@@ -1258,6 +1258,7 @@ pub(crate) fn lowlevel_borrow_signature(op: LowLevel) -> &'static [Ownership] {
         StrReleaseExcessCapacity => &[OWNED],
         ListIncref => &[OWNED],
         ListDecref => &[OWNED],
+        StrWithAsciiLowercased => &[OWNED],
 
         Eq | NotEq => &[BORROWED, BORROWED],
 

crates/compiler/solve/tests/solve_expr.rs

@@ -3824,6 +3824,18 @@ mod solve_expr {
         );
     }
 
+    #[test]
+    fn str_with_ascii_lowercased() {
+        infer_eq_without_problem(
+            indoc!(
+                r"
+                Str.with_ascii_lowercased
+                "
+            ),
+            "Str -> Str",
+        );
+    }
+
     #[test]
     fn list_take_first() {
         infer_eq_without_problem(

crates/compiler/test_gen/src/gen_str.rs

@@ -2060,3 +2060,29 @@ fn str_drop_suffix() {
         RocStr
     );
 }
+
+#[test]
+#[cfg(any(feature = "gen-llvm", feature = "gen-dev", feature = "gen-wasm"))]
+fn with_ascii_lowercased() {
+    assert_evals_to!(
+        r#"
+        Str.with_ascii_lowercased "cOFFÉ"
+        "#,
+        RocStr::from("coffÉ"),
+        RocStr
+    );
+}
+
+#[test]
+#[cfg(any(feature = "gen-llvm", feature = "gen-dev", feature = "gen-wasm"))]
+fn with_ascii_lowercased_non_zero_refcount() {
+    assert_evals_to!(
+        r#"
+        original = "cOFFÉ cOFFÉ cOFFÉ cOFFÉ cOFFÉ cOFFÉ cOFFÉ cOFFÉ cOFFÉ cOFFÉ cOFFÉ cOFFÉ"
+        res = Str.with_ascii_lowercased original
+        Str.dropPrefix res original
+        "#,
+        RocStr::from("coffÉ coffÉ coffÉ coffÉ coffÉ coffÉ coffÉ coffÉ coffÉ coffÉ coffÉ coffÉ"),
+        RocStr
+    );
+}