[pull] master from GaijinEntertainment:master

.github/workflows/build.yml

@@ -256,6 +256,15 @@ jobs:
             ;;
         esac
 
+    # sccache on PRs: per-TU content-hash cache via GHA backend. Master clean.
+    - if: github.event_name == 'pull_request'
+      uses: mozilla-actions/sccache-action@v0.0.10
+    - if: github.event_name == 'pull_request'
+      run: |
+        echo "SCCACHE_GHA_ENABLED=true" >> $GITHUB_ENV
+        echo "CMAKE_C_COMPILER_LAUNCHER=sccache" >> $GITHUB_ENV
+        echo "CMAKE_CXX_COMPILER_LAUNCHER=sccache" >> $GITHUB_ENV
+
     - name: "Build: Daslang"
       run: |
         set -eux

.github/workflows/extended_checks.yml

@@ -124,6 +124,15 @@ jobs:
         ./vcpkg/vcpkg install openssl:x64-windows --binarycaching
         echo "VCPKG_ROOT=$(pwd)/vcpkg" >> $GITHUB_ENV
 
+    # sccache on PRs: per-TU content-hash cache via GHA backend. Master clean.
+    - if: github.event_name == 'pull_request'
+      uses: mozilla-actions/sccache-action@v0.0.10
+    - if: github.event_name == 'pull_request'
+      run: |
+        echo "SCCACHE_GHA_ENABLED=true" >> $GITHUB_ENV
+        echo "CMAKE_C_COMPILER_LAUNCHER=sccache" >> $GITHUB_ENV
+        echo "CMAKE_CXX_COMPILER_LAUNCHER=sccache" >> $GITHUB_ENV
+
     - name: "Build: Daslang (Release)"
       run: |
         set -eux

.github/workflows/pages.yml

@@ -73,23 +73,22 @@ jobs:
         working_directory: build/latex
         continue_on_error: true
 
+    - name: Setup Emscripten
+      uses: emscripten-core/setup-emsdk@v11
+      with:
+        # Pinned: 5.0.7's clang crashes on utils/dasFormatter/ds_parser.cpp
+        # in diagnostic snippet rendering. 5.0.3 known-good.
+        version: '5.0.3'
+
     - name: "Build WASM (Emscripten) for /playground/ + /files/wasm/"
       run: |
         set -eux
         cd web
-        # step0 only clones the emsdk wrapper — the toolchain itself
-        # (emsdk/upstream/emscripten/) lands via the install+activate pair
-        # that web/step1_emsdk_activate_linux.sh does locally. Without these,
-        # cmake's -DCMAKE_TOOLCHAIN_FILE points at a path that doesn't
-        # exist yet and the WASM build silently fails under continue-on-error.
-        bash step0_emsdk_install.sh
-        ./emsdk/emsdk install latest
-        ./emsdk/emsdk activate latest
-        bash -c "source emsdk/emsdk_env.sh && \
-                 mkdir -p build && cd build && \
-                 cmake -DCMAKE_BUILD_TYPE=Release -G Ninja \
-                       -DCMAKE_TOOLCHAIN_FILE=../emsdk/upstream/emscripten/cmake/Modules/Platform/Emscripten.cmake ../ && \
-                 ninja"
+        mkdir -p build && cd build
+        emcmake cmake -DCMAKE_BUILD_TYPE=Release -G Ninja ..
+        # daslang_static.wasm playground + stage_site_playground (vendored UI
+        # + js/wasm) + stage_site_playground_wasm (cross-compiled examples).
+        ninja daslang_static stage_site_playground stage_site_playground_wasm
       continue-on-error: true
 
     - name: "Fetch dasProfile benchmark JSON"

.github/workflows/playground-e2e.yml

@@ -8,13 +8,13 @@ on:
   pull_request:
     paths:
       - 'site/**'
-      - 'web/ui/**'
+      - 'web/examples/ui/**'
       - '.github/workflows/playground-e2e.yml'
   push:
     branches-ignore: [master]
     paths:
       - 'site/**'
-      - 'web/ui/**'
+      - 'web/examples/ui/**'
       - '.github/workflows/playground-e2e.yml'
   workflow_dispatch:
 
@@ -48,9 +48,9 @@ jobs:
           cp site/downloads.html _site/ || true
           cp -r site/files _site/files
 
-          # Playground: vendor web/ui IDE (no WASM artifacts)
-          cp -r web/ui/src/* _site/playground/
-          cp -r web/ui/samples _site/playground/samples 2>/dev/null || true
+          # Playground: vendor web/examples/ui IDE (no WASM artifacts)
+          cp -r web/examples/ui/src/* _site/playground/
+          cp -r web/examples/ui/samples _site/playground/samples 2>/dev/null || true
           cp site/playground/index.html _site/playground/index.html
           cp site/playground/forge-skin.css _site/playground/
           # Forge playground scripts (init shim + tabs + share + splitter).

.github/workflows/wasm_build.yml

@@ -108,6 +108,17 @@ jobs:
             ./emsdk/emsdk_env.bat
             ;;
           *)
+            # macOS stock bison is 2.3; install Homebrew bison 3.x and shadow
+            # the system one on PATH so any FIND_FLEX_AND_BISON pickup is sane.
+            if [[ "$(uname -m)" == "arm64" ]]; then
+              brew install bison
+              echo 'export PATH="/opt/homebrew/opt/bison/bin:$PATH"' >> ~/.bash_profile
+              export PATH="/opt/homebrew/opt/bison/bin:$PATH"
+            else
+              brew install bison
+              echo 'export PATH="/usr/local/opt/bison/bin:$PATH"' >> ~/.bash_profile
+              export PATH="/usr/local/opt/bison/bin:$PATH"
+            fi
             ./emsdk/emsdk install latest
             ./emsdk/emsdk activate latest
             source "/Users/runner/work/daScript/daScript/emsdk/emsdk_env.sh"
@@ -117,7 +128,6 @@ jobs:
       run: |
         set -eux
         cd web
-        cp ../CMakeXxdImpl.txt .
         rm -r -f cmake_temp
         mkdir cmake_temp
         cd cmake_temp
@@ -126,5 +136,58 @@ jobs:
 
     - name: "Test: hello world via Node.js"
       run: |
+        set -eux
+        # Use emsdk-bundled Node (deterministic version that supports the
+        # --experimental-wasm-exnref flag). System Node may be older or use
+        # a different spelling for the EH proposal flag.
+        source ./emsdk/emsdk_env.sh
+        cd web
+        # Modern wasm EH proposal (try_table / exnref) is gated in Node 22.
+        # Default-on in Node 24+. Force-enable for forward compat with current LTS.
+        "$EMSDK_NODE" --experimental-wasm-exnref test/dastest_wasm.js ../ ./output
+
+  ###########################################################
+  wasm_cross:
+  ###########################################################
+    # Cross-compile utility mains to wasm32 via dasLLVM and execute under
+    # wasmtime. Smoke-tests the runtime-linked cross-compile pipeline
+    # end-to-end (see modules/dasLLVM/README.md "Cross-compilation").
+    needs: pre_job
+    if: needs.pre_job.outputs.should_skip != 'true'
+    runs-on: ubuntu-latest-fat
+    steps:
+    - name: "SCM Checkout"
+      uses: actions/checkout@v4
+
+    - uses: lukka/get-cmake@latest
+
+    - name: Setup Emscripten
+      uses: emscripten-core/setup-emsdk@v11
+      with:
+        # 5.0.7 ships clang commit 7b58716d that crashes in
+        # TextDiagnostic::emitSnippetAndCaret on utils/dasFormatter/ds_parser.cpp
+        # (std::length_error during diagnostic snippet rendering). 5.0.3 works.
+        version: '5.0.3'
+
+    - name: Setup Wasmtime
+      uses: bytecodealliance/actions/wasmtime/setup@v1
+      with:
+        version: "44.0.1"
+
+    - name: "Install: graphics dev libs"
+      run: |
+        set -eux
+        sudo apt-get update -y
+        # OpenGL + GLFW headers needed by dasGlfw (enabled by default in host
+        # cmake). Same set as extended_checks.yml.
+        sudo apt-get install --no-install-recommends -y \
+          mesa-common-dev libglu1-mesa-dev libgl1-mesa-dev libglfw3-dev \
+          libx11-dev libxrandr-dev libxcursor-dev libxinerama-dev libxi-dev
+
+    - name: "Build + run cross-compiled wasm examples"
+      run: |
+        set -eux
         cd web
-        node test/dastest_wasm.js ../ ./output
+        mkdir -p cmake_temp && cd cmake_temp
+        emcmake cmake -DCMAKE_BUILD_TYPE=Release -G Ninja ..
+        ninja run_wasm_examples

CLAUDE.md

@@ -263,6 +263,7 @@ Full migration table (when reading older docs that say `var inscope` or `<-` for
 | `unsafe(x + y)` / `unsafe { let d = x + y }` where nothing inside requires unsafe | drop the wrap | STYLE024: redundant `unsafe` — flagged when no descendant matches a known inherently-unsafe shape (reinterpret/upcast cast, `delete`, `addr`, table-index, variant-write, ExprCallFunc with `unsafeOperation`). Macro-generated subtrees (`genFlags.generated == true`) skipped per design |
 | `unsafe { stmt1; stmt2; stmt_needing_unsafe }` (only ONE stmt actually needs unsafe) | `stmt1; stmt2; unsafe(<sub-expr>)` | STYLE025: narrow block-form unsafe to expression-form on the single unsafe-needing statement. Silent when ≥2 statements need unsafe (block is justified) |
 | `unsafe { ...; unsafe { ... }; ... }` (nested `unsafe { }` block) | drop the inner wrap | STYLE026: outer `unsafe` already covers the whole inner scope, so the inner block is pure noise. Closure / lambda / generator bodies are NOT nested for this rule — they execute in a separate context where the outer wrap does not propagate |
+| `for (s in A) { B \|> push(s) }` / `push_clone(s)` (iter-var only) | `B \|> push_from(A)` / `push_clone_from(A)` | PERF022: the bulk overload in builtin.das reserves combined capacity up front. Single name `push`/`push_clone` is overloaded between single-element and bulk (ambiguous when destination is `array<T[]>`); the `_from` suffix names the bulk intent. Source must be `array<T>` or C-array — range/iterator sources are not flagged. `emplace` is out of scope (const iter-var can't be moved) |
 
 For path/filename ops use `fio` helpers (`base_name`/`dir_name`/`path_join`/etc.) — see `skills/filesystem.md`. Never hand-roll `rfind("/")` / slice — misses Windows separators.
 

benchmarks/sql/M4_DECS_EXPANSION.md

@@ -337,3 +337,21 @@ Two new planners `plan_decs_order_family` (mirrors `plan_order_family`) and `pla
 All 6 rows improve 1.5-2.7× (avg ~2.1×). m4 lands within 2× of m3f on most rows — the remaining gap is the Wave 4 multi-component `get_ro` overhead (sort dominates the wall-clock so the gap doesn't fully close even when the splice fires). `bare_order_where` is the tightest squeeze (1.5×) because sort over 100K rows is the bottleneck; once the m4 buffer build matches m3f (~117 ns), the rest is pure sort time.
 
 **Coverage:** bare `_order_by` + to_array, bare `_order_by_descending` + to_array, `_order_by` + take, `_order_by_descending` + take, `_order_by` + first, `_order_by_descending` + first, `_order_by` + first_or_default (nonempty + empty), `_where + _order_by + take`, `_where + _order_by + first`; `.reverse() + to_array`, `.reverse() + take(N)` (in-range, beyond-length, and zero — last uses runtime-var to defeat const-fold of PERF017 false-positive on splice-emitted `0 < length(buf)`), `.reverse() + count` (no-buffer counter loop), `_where + .reverse() + first_or_default` (empty + nonempty), `_where + .reverse() + to_array`, `_where + _select + .reverse() + to_array` (projection through reverse). AST shape gates for: order+take (top_n_by emit + no to_sequence + 1 for_each_archetype), order+first (min_by emit + panic-on-empty guard + no top_n_by), reverse+to_array (1 reverse_inplace), reverse+count (no decs_buf, no reverse_inplace). +20 tests (100 → 120 in file).
+
+## Update — Slice 5c distinct/distinct_by on decs (2026-05-21, plan_decs_distinct)
+
+New `plan_decs_distinct` planner mirrors `plan_distinct` — streaming dedup via `var inscope seen : table<typedecl(_::unique_key(...))>` hoisted above `for_each_archetype` so the seen-table persists across archetypes. Inserted in the cascade BEFORE `plan_distinct` (same rule as 5d/5e — decs bridge match is stricter). Two emit shapes:
+
+- **Buffer-required** (`to_array` default, optionally `take(N)`): per-element splice computes the key, `key_exists` check, on miss does `taken++ → seen|>insert → buf|>push_clone`. When `take(N)` is present, outer iteration switches to `for_each_archetype_find` and the take-cap `return true` propagates across archetypes — true streaming early-exit. Take-limit bound to a `let` at outer scope so a side-effecting `take(arg)` evaluates exactly once.
+- **Buffer-elided** (`count`/`long_count` → `length(seen)`/`int64(length(seen))`; `sum` folds inline at fresh-key site via `acc += <projection>`): no buffer allocation, only the seen-table is materialized. count + take both present cascades to tier 2 (matches array-side).
+
+Side-effecting `_select(proj)` upstream binds once per element to a fresh `decs_pv` local — key + buf push (or sum fold) share the bind, matching array-side single-eval per source element. `distinct_by(key)` wraps the key block in `invoke(<keyBlock>, <projection-or-tup>)` and the seen-table's value type is derived via `_::unique_key(invoke(<keyBlock>, default<elemT>))` so the table key type tracks the key function's return type.
+
+| benchmark | shape | m1 sql | m3 | m3f (array splice) | m4 (eager bridge, was) | m4 (Slice 5c, now) | m4 win |
+|---|---|---:|---:|---:|---:|---:|---:|
+| distinct_count | `_select(_.brand).distinct().to_array()` | 41 | 44 | 15 | 97 | **28** | 3.5× |
+| distinct_take  | `_select(_.brand).distinct().take(3).to_array()` | 0 | 30 | 0 | 34 | **0** | matches m3f |
+
+`distinct_count` lands at 28 ns/op vs m3f's 15 — the ~13 ns gap is the Wave 4 multi-component `get_ro` floor (3 components participate in the inner for-loop even when chain only reads `brand`). `distinct_take` collapses to 0 ns/op — early-exit at the 3rd distinct brand visits only ~3 source elements regardless of N=100K, same as the array-side splice.
+
+**Coverage:** `_select(_.brand).distinct()` + to_array / count / long_count / sum / take(N) / take(0) / take(N>num_distinct), `_where(_.id<8)._select(_.brand).distinct()`, `_distinct_by(_.brand)` + to_array / count / take(N), empty-decs distinct yields empty, side-effecting take(N) arg evaluates exactly once at invoke entry. AST shape gates for: distinct+count (no to_sequence, no decs_buf, single key_exists, plain for_each_archetype), distinct+take (for_each_archetype_find + decs_buf + decs_seen + decs_taken counters), distinct_by+to_array (unique_key wrapping the key invocation), distinct+sum (no decs_buf, decs_acc declared+folded). +17 tests (122 → 139 in file).

daslib/algorithm.das

@@ -54,13 +54,9 @@ def reverse(var a : array<auto>) {
 def combine(a, b : array<auto(TT)>) {
     //! Returns a new array containing elements from a followed by b.
     var c : array<TT>
-    reserve(c, length(a) + length(b))
-    for (t in a) {
-        c |> push_clone(t)
-    }
-    for (t in b) {
-        c |> push_clone(t)
-    }
+    c |> reserve(long_length(a) + long_length(b))
+    c |> push_clone_from(a)
+    c |> push_clone_from(b)
     return <- c
 }