process and fs; Array flat/flatMap/toSpliced, Map.groupBy, String.fromCodePoint

.gitignore

@@ -11,4 +11,5 @@
 *.lib
 *.pdb
 *.log
-__jit/
+__jit/
+__fs_test/

README.md

@@ -133,6 +133,7 @@ A test that reads standard input (`console.readLine`) gets it from `tests/<name>
 ## Docs
 
 - [fetch / Headers / Response](docs/fetch.md) — built-in HTTP client
+- [process / fs](docs/process-fs.md) — the command line, the environment and the file system
 
 ## Implemented classes & functions
 
@@ -145,7 +146,7 @@ Globals (`src/lib.ts`):
 - `Date`
 - `RegExp`, `MatchResults`, `MatchIndicesResults`
 - `String`: `String.fromCharCode` (each code written as UTF-8, a surrogate pair as one
-  character), plus the `string` prototype methods it wraps: `at`, `charAt`, `charCodeAt`,
+  character), `String.fromCodePoint` (each code point as UTF-8; outside 0 to 0x10FFFF a `RangeError`), plus the `string` prototype methods it wraps: `at`, `charAt`, `charCodeAt`,
   `codePointAt`, `concat`, `endsWith`, `includes`, `indexOf`, `lastIndexOf`,
   `localeCompare`, `match`, `matchAll`, `normalize`, `padEnd`, `padStart`, `repeat`,
   `replace`, `replaceAll`, `search`, `slice`, `split`, `startsWith`, `substring`,
@@ -192,7 +193,10 @@ Generics (`src/generics/lib.generics.ts`):
   `Int16Array`, `Uint16Array`, `Int32Array`, `Uint32Array`, `BigInt64Array`,
   `BigUint64Array`, `Float32Array`, `Float64Array`). A typed array is made from an array of
   its elements (`new Uint8Array([1, 2, 3])`) or a length (`new Uint8Array(4096)`, each element 0)
-- `Map<K, V>`
+- `T[]` and `Array<T>` also have `flatMap`, `toSpliced` (`splice` on a copy), and `T[][]` has
+  `flat()`, which takes away one level only: a depth would make the result's type depend on a number
+- `Map<K, V>`, and `Map.groupBy(items, keySelector)`, a `Map` of each key to its items, the keys in
+  the order they first come (not under `-mm=own`, whose Map values cannot be arrays)
 - `Set<V>`
 - `Atomics`: `add`, `sub`, `and`, `or`, `xor`, `exchange`, `compareExchange`, `load`,
   `store`, `isLockFree` over the integer typed arrays, and `wait`/`notify` over an
@@ -290,6 +294,25 @@ Generics (`src/generics/lib.generics.ts`):
   const text = new TextDecoder().decode(Stdio.readAll());
   Stdio.write(new TextEncoder().encode(text.toUpperCase()));
   ```
+- `process`, as Node's, with functions where Node has properties: `argv()`, `getEnv(name)` (or
+  `undefined`), `setEnv`, `unsetEnv`, `env()` (`[name, value]` pairs), `cwd()`, `chdir`, `pid()`, `platform()`,
+  `exit(code)`. See [docs/process-fs.md](docs/process-fs.md).
+- `fs`, synchronous, by Node's names: `existsSync`, `statSync`, `readFileSync` (a `Uint8Array`),
+  `readTextFileSync` (a string), `writeFileSync`, `writeTextFileSync`, `appendFileSync`,
+  `appendTextFileSync`, `readdirSync` (sorted), `mkdirSync` (`{ recursive }`), `rmSync`
+  (`{ recursive, force }`), `rmdirSync`, `renameSync`, `copyFileSync`. An error is an `Error` with
+  Node's message, `"ENOENT: no such file or directory, open 'a.txt'"`. See
+  [docs/process-fs.md](docs/process-fs.md).
+
+  ```ts
+  const args = process.argv();
+  const name = args[args.length - 1];
+  if (!fs.existsSync(name)) {
+      console.error(`${name}: not found`);
+      process.exit(1);
+  }
+  fs.writeTextFileSync(name + ".upper", fs.readTextFileSync(name).toUpperCase());
+  ```
 
 Errors (`src/core/core.d.ts`):
 

docs/process-fs.md

@@ -0,0 +1,103 @@
+# process / fs
+
+The command line, the environment, the working directory and the file system, synchronously. Both
+are namespaces in `src/generics/lib.generics.ts` over the C++ wrapper `src/wrappers/os.cpp`.
+
+## process
+
+Node's `process`, with functions where Node has properties (a namespace in the library holds no
+state of its own, so `argv` and `env` are read when they are asked for):
+
+```ts
+namespace process {
+    function argv(): string[];                       // Node: process.argv
+    function getEnv(name: string): string | undefined; // Node: process.env[name]
+    function setEnv(name: string, value: string): void;
+    function unsetEnv(name: string): void;           // Node: delete process.env[name]
+    function env(): [string, string][];              // every variable, [name, value]
+    function cwd(): string;
+    function chdir(directory: string): void;
+    function pid(): int;
+    function platform(): string;                     // "win32", "linux", "android", "darwin"
+    function exit(code?: int): void;
+}
+```
+
+- `argv()[0]` is the program. A compiled program's arguments follow it. Under the JIT the program
+  is `tslang`, so its options come next, then the script, then the script's own arguments - the way
+  `node script.js a b` gives `["node", "script.js", "a", "b"]`. On Linux and Android the arguments
+  are read from `/proc/self/cmdline`, on Windows from `GetCommandLineW`.
+- `setEnv` and `unsetEnv` change the environment of this process and of the processes it starts.
+  A name that is empty or has an `=` in it is an `Error` (`EINVAL`). Unsetting a variable that is
+  not set is no error. On Windows the variables are read and written through the Win32 API, which
+  every C runtime in the process shares (under the JIT this library has a C runtime of its own),
+  and names are case-insensitive there, as always on Windows.
+- `exit(code)` flushes every C stream and ends the process: nothing after it runs, no `finally`
+  block either. Under the JIT on Windows it flushes this library's streams (`console.log`,
+  `Stdio`), not the host's, so text from `print()` not yet flushed may be lost there.
+
+## fs
+
+```ts
+namespace fs {
+    function existsSync(path: string): boolean;
+    function statSync(path: string): FileStats;   // { isFile, isDirectory, size, mtimeMs }
+    function readFileSync(path: string): Uint8Array;
+    function readTextFileSync(path: string): string;
+    function writeFileSync(path: string, data: Uint8Array): void;
+    function writeTextFileSync(path: string, text: string): void;
+    function appendFileSync(path: string, data: Uint8Array): void;
+    function appendTextFileSync(path: string, text: string): void;
+    function readdirSync(path: string): string[];
+    function mkdirSync(path: string, options?: { recursive?: boolean }): void;
+    function rmSync(path: string, options?: { recursive?: boolean, force?: boolean }): void;
+    function rmdirSync(path: string): void;
+    function renameSync(oldPath: string, newPath: string): void;
+    function copyFileSync(src: string, dest: string): void;
+}
+```
+
+- The names are Node's. Node's `readFileSync(path, "utf8")` and its `writeFileSync` that takes a
+  string are `readTextFileSync` and `writeTextFileSync` here (Deno's names), since a function here
+  has one result type: `readFileSync` gives bytes, `writeFileSync` takes them.
+- Text is UTF-8, and a string ends at its first 0, so a file that is not text is read with
+  `readFileSync`. Paths are UTF-8 on every platform; on Windows they go through the wide API, so a
+  name outside the ANSI code page works.
+- `statSync` follows symbolic links. `size` is in bytes; `mtimeMs` is the last modification in
+  milliseconds since 1970, as `new Date(ms)` takes it (to the second on Windows).
+- `readdirSync` gives the names in the directory, without `.` and `..`, sorted: the operating system
+  gives them in no set order.
+- `mkdirSync` without `recursive` needs the parent, and a directory already there is `EEXIST`; with
+  it every missing parent is made too and a directory already there is no error.
+- `rmSync` removes a file; a directory only with `recursive`, and then all it holds (`EISDIR`
+  without). With `force` a path that is not there is no error. `rmdirSync` removes an empty
+  directory.
+- `renameSync` and `copyFileSync` replace a file already at the destination.
+- Files of 2 GB or more are an `EFBIG` error: lengths are an `int`, as `Stdio`'s are.
+
+An error is an `Error` whose message is Node's: the code, what it means, the call and the path.
+
+```ts
+try {
+    fs.readTextFileSync("settings.json");
+} catch (e: Error) {
+    console.log(e.message); // ENOENT: no such file or directory, open 'settings.json'
+}
+```
+
+The code is one of `ENOENT`, `EEXIST`, `EACCES`, `EPERM`, `ENOTDIR`, `EISDIR`, `ENOTEMPTY`, `EBUSY`,
+`EINVAL`, `ENAMETOOLONG`, `EXDEV`, `EFBIG`, `ENOSPC`, `EROFS`, `EMFILE`, or `EIO` for any other; what
+it means is the C runtime's wording, so its letter case differs between platforms.
+
+## Example
+
+```ts
+function main() {
+    const args = process.argv();
+    const dir = args.length > 1 ? args[args.length - 1] : process.cwd();
+    for (const name of fs.readdirSync(dir)) {
+        const stats = fs.statSync(dir + "/" + name);
+        console.log(stats.isDirectory ? `${name}/` : `${name} ${stats.size}`);
+    }
+}
+```

scripts/build.sh

@@ -87,12 +87,13 @@ $TOOL $DBG_GCC $CPP_FLAGS -c $SRC/src/wrappers/regex.cpp -o $OUTPUT/$LIB_OUT/reg
 $TOOL $DBG_GCC $CPP_FLAGS $WRAPPER_MM_DEF -c $SRC/src/wrappers/thread.cpp -o $OUTPUT/$LIB_OUT/thread.o
 $TOOL $DBG_GCC $CPP_FLAGS -c $SRC/src/wrappers/http_linux.cpp -o $OUTPUT/$LIB_OUT/http_linux.o
 $TOOL $DBG_GCC $CPP_FLAGS -c $SRC/src/wrappers/memory.cpp -o $OUTPUT/$LIB_OUT/memory.o
+$TOOL $DBG_GCC $CPP_FLAGS -c $SRC/src/wrappers/os.cpp -o $OUTPUT/$LIB_OUT/os.o
 
 $BIN_PATH/$TOOL_NAME $DBG_OPTS $MM_OPT --emit=obj --export=none --nowarn --no-default-lib $SRC/src/lib.linux.ts $PIC -o $OUTPUT/$LIB_OUT/lib.linux.o
 
 # Build Lib
 $BIN_PATH/$TOOL_NAME $DBG_OPTS $MM_OPT --emit=obj --export=none --nowarn --no-default-lib $SRC/src/lib.ts $PIC -o $OUTPUT/$LIB_OUT/lib.o
-$ARC rcs $OUTPUT/$LIB_OUT/libTypeScriptDefaultLibCore.a $OUTPUT/$LIB_OUT/lib.o $OUTPUT/$LIB_OUT/lib.linux.o $OUTPUT/$LIB_OUT/io.o $OUTPUT/$LIB_OUT/datetime.o $OUTPUT/$LIB_OUT/regex.o $OUTPUT/$LIB_OUT/thread.o $OUTPUT/$LIB_OUT/http_linux.o $OUTPUT/$LIB_OUT/memory.o
+$ARC rcs $OUTPUT/$LIB_OUT/libTypeScriptDefaultLibCore.a $OUTPUT/$LIB_OUT/lib.o $OUTPUT/$LIB_OUT/lib.linux.o $OUTPUT/$LIB_OUT/io.o $OUTPUT/$LIB_OUT/datetime.o $OUTPUT/$LIB_OUT/regex.o $OUTPUT/$LIB_OUT/thread.o $OUTPUT/$LIB_OUT/http_linux.o $OUTPUT/$LIB_OUT/memory.o $OUTPUT/$LIB_OUT/os.o
 
 # A static archive cannot record its own dependencies on ELF (there is no ld.bfd equivalent of
 # the `#pragma comment(lib, ...)` http.cpp uses on Windows), so programs linking
@@ -110,7 +111,7 @@ EOF
 # undefined on purpose: they must bind to the one collector already in the process
 # (libTypeScriptRuntime.so under the JIT, libgc.a in an exe); linking libgc.a in here would
 # give the library a second, separate heap.
-g++ -shared $DBG_GCC $OUTPUT/$LIB_OUT/lib.o $OUTPUT/$LIB_OUT/lib.linux.o $OUTPUT/$LIB_OUT/io.o $OUTPUT/$LIB_OUT/datetime.o $OUTPUT/$LIB_OUT/regex.o $OUTPUT/$LIB_OUT/thread.o $OUTPUT/$LIB_OUT/http_linux.o $OUTPUT/$LIB_OUT/memory.o -lcurl -lm -o $OUTPUT/$DLL_OUT/libTypeScriptDefaultLib.so
+g++ -shared $DBG_GCC $OUTPUT/$LIB_OUT/lib.o $OUTPUT/$LIB_OUT/lib.linux.o $OUTPUT/$LIB_OUT/io.o $OUTPUT/$LIB_OUT/datetime.o $OUTPUT/$LIB_OUT/regex.o $OUTPUT/$LIB_OUT/thread.o $OUTPUT/$LIB_OUT/http_linux.o $OUTPUT/$LIB_OUT/memory.o $OUTPUT/$LIB_OUT/os.o -lcurl -lm -o $OUTPUT/$DLL_OUT/libTypeScriptDefaultLib.so
 
 # Copy
 # Stage into a single shared defaultlib tree with per-target, per-build and per-model subfolders

scripts/build_android.bat

@@ -120,14 +120,14 @@ md "%OBJ%" "%LIB_OUT%"
 rem thread.cpp registers the threads it starts with the collector, which only the gc build links
 set WRAPPER_MM_DEF=
 if "%MM%"=="gc" set WRAPPER_MM_DEF=-DTSLANG_MM_GC
-for %%w in (io datetime regex thread http_stub memory) do (
+for %%w in (io datetime regex thread http_stub memory os) do (
     %CXX% %WRAPPER_MM_DEF% -c "%ROOT%\src\wrappers\%%w.cpp" -o "%OBJ%\%%w.o" || exit /b 1
 )
 
 %TSC% "%ROOT%\src\lib.linux.ts" -o "%OBJ%\lib.linux.o" || exit /b 1
 %TSC% "%ROOT%\src\lib.ts" -o "%OBJ%\lib.o" || exit /b 1
 
-set OBJS="%OBJ%\lib.o" "%OBJ%\lib.linux.o" "%OBJ%\io.o" "%OBJ%\datetime.o" "%OBJ%\regex.o" "%OBJ%\thread.o" "%OBJ%\http_stub.o" "%OBJ%\memory.o"
+set OBJS="%OBJ%\lib.o" "%OBJ%\lib.linux.o" "%OBJ%\io.o" "%OBJ%\datetime.o" "%OBJ%\regex.o" "%OBJ%\thread.o" "%OBJ%\http_stub.o" "%OBJ%\memory.o" "%OBJ%\os.o"
 %AR% rcs "%LIB_OUT%\libTypeScriptDefaultLib.a" %OBJS% || exit /b 1
 
 rd /s /q "%OBJ%"

scripts/build_core.bat

@@ -225,22 +225,23 @@ rem Build native wrappers for C++ code
 %TSLANG_CC% %DBG_CL% /EHsc /Wall %WRAPPER_MM_DEF% /c /Fo%OUTPUT%\%LIB_OUT%\ %SRC%\src\wrappers\thread.cpp
 %TSLANG_CC% %DBG_CL% /EHsc /Wall /c /Fo%OUTPUT%\%LIB_OUT%\ %SRC%\src\wrappers\http.cpp
 %TSLANG_CC% %DBG_CL% /EHsc /Wall /c /Fo%OUTPUT%\%LIB_OUT%\ %SRC%\src\wrappers\memory.cpp
+%TSLANG_CC% %DBG_CL% /EHsc /Wall /c /Fo%OUTPUT%\%LIB_OUT%\ %SRC%\src\wrappers\os.cpp
 
 rem Build OS-specific Lib
 echo Build OS-specific Lib
 %TOOL_PATH%\%TOOL_NAME%.exe %TRIPLE_OPT% %DBG% %MM_OPT% --emit=obj --export=none --nowarn --no-default-lib %SRC%\src\lib.win32.ts -o %OUTPUT%\%LIB_OUT%\lib.win32.obj
 
 rem Build DLL
 echo Build DLL
-%TOOL_PATH%\%TOOL_NAME%.exe %TRIPLE_OPT% %DBG% %MM_OPT% --emit=dll --gc-lib-path=%GC_SHARED_LIB_PATH% %TSLANG_RUNTIME_OPT% --embed-declarations=false --nowarn --no-default-lib %SRC%\src\lib.ts --obj=%OUTPUT%\%LIB_OUT%\lib.win32.obj --obj=%OUTPUT%\%LIB_OUT%\io.obj --obj=%OUTPUT%\%LIB_OUT%\datetime.obj --obj=%OUTPUT%\%LIB_OUT%\regex.obj --obj=%OUTPUT%\%LIB_OUT%\thread.obj --obj=%OUTPUT%\%LIB_OUT%\http.obj --obj=%OUTPUT%\%LIB_OUT%\memory.obj -o %OUTPUT%\%DLL_OUT%\TypeScriptDefaultLib.dll
+%TOOL_PATH%\%TOOL_NAME%.exe %TRIPLE_OPT% %DBG% %MM_OPT% --emit=dll --gc-lib-path=%GC_SHARED_LIB_PATH% %TSLANG_RUNTIME_OPT% --embed-declarations=false --nowarn --no-default-lib %SRC%\src\lib.ts --obj=%OUTPUT%\%LIB_OUT%\lib.win32.obj --obj=%OUTPUT%\%LIB_OUT%\io.obj --obj=%OUTPUT%\%LIB_OUT%\datetime.obj --obj=%OUTPUT%\%LIB_OUT%\regex.obj --obj=%OUTPUT%\%LIB_OUT%\thread.obj --obj=%OUTPUT%\%LIB_OUT%\http.obj --obj=%OUTPUT%\%LIB_OUT%\memory.obj --obj=%OUTPUT%\%LIB_OUT%\os.obj -o %OUTPUT%\%DLL_OUT%\TypeScriptDefaultLib.dll
 
 rem Build Lib
 echo Build Lib
 %TOOL_PATH%\%TOOL_NAME%.exe %TRIPLE_OPT% %DBG% %MM_OPT% --emit=obj --export=none --nowarn --no-default-lib %SRC%\src\lib.ts -o %OUTPUT%\%LIB_OUT%\lib.obj
 rem %TOOL_PATH%\%TOOL_NAME%.exe %TRIPLE_OPT% %DBG% %MM_OPT% --emit=llvm --export=none %SRC%\src\lib.ts -o %OUTPUT%\%LIB_OUT%\lib.ll
 rem %TOOL_PATH%\%TOOL_NAME%.exe %TRIPLE_OPT% %DBG% %MM_OPT% --emit=mlir --export=none %SRC%\src\lib.ts 2> %OUTPUT%\%LIB_OUT%\lib.mlir
 
-%TSLANG_AR% /out:%OUTPUT%\%LIB_OUT%\TypeScriptDefaultLib.lib %OUTPUT%\%LIB_OUT%\lib.obj %OUTPUT%\%LIB_OUT%\lib.win32.obj %OUTPUT%\%LIB_OUT%\io.obj %OUTPUT%\%LIB_OUT%\datetime.obj %OUTPUT%\%LIB_OUT%\regex.obj %OUTPUT%\%LIB_OUT%\thread.obj %OUTPUT%\%LIB_OUT%\http.obj %OUTPUT%\%LIB_OUT%\memory.obj
+%TSLANG_AR% /out:%OUTPUT%\%LIB_OUT%\TypeScriptDefaultLib.lib %OUTPUT%\%LIB_OUT%\lib.obj %OUTPUT%\%LIB_OUT%\lib.win32.obj %OUTPUT%\%LIB_OUT%\io.obj %OUTPUT%\%LIB_OUT%\datetime.obj %OUTPUT%\%LIB_OUT%\regex.obj %OUTPUT%\%LIB_OUT%\thread.obj %OUTPUT%\%LIB_OUT%\http.obj %OUTPUT%\%LIB_OUT%\memory.obj %OUTPUT%\%LIB_OUT%\os.obj
 
 del %OUTPUT%\%LIB_OUT%\lib.obj
 del %OUTPUT%\%LIB_OUT%\lib.win32.obj
@@ -250,6 +251,7 @@ del %OUTPUT%\%LIB_OUT%\regex.obj
 del %OUTPUT%\%LIB_OUT%\thread.obj
 del %OUTPUT%\%LIB_OUT%\http.obj
 del %OUTPUT%\%LIB_OUT%\memory.obj
+del %OUTPUT%\%LIB_OUT%\os.obj
 
 rem Stage into a single shared defaultlib tree with per-target, per-build, per-model subfolders
 rem under dll\ and lib\. Only the current build's subfolders are refreshed so the other