Next release: timely 0.32, columnar and tactic work, interactive/ and ddir

.github/workflows/test.yml

@@ -39,11 +39,77 @@ jobs:
         run: cargo test --workspace --all-targets
       - name: Cargo doc test
         run: cargo test --doc
+
+  # The server-backed suites and the explanation sweeps, once, against a release build: they
+  # check answers, not timings, and a debug server made them most of the suite's time. The
+  # matrix above keeps the debug build's assertions through `cargo test`.
+  ldbc:
+    name: LDBC and explanation suites (release)
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+      - uses: actions-rust-lang/setup-rust-toolchain@v1
+      - name: Build the release server
+        run: cargo build --release -p ddir-server
+      # One small server-backed workload; no data download or timing thresholds.
+      - name: LDBC-derived correctness smoke
+        run: |
+          for workers in 1 4; do
+            python3 interactive/bench/ldbc/run.py --server target/release/ddir_server \
+              --workers "$workers" --rounds 1 --warmup 0 --batch-size 9 --changes 1
+          done
+      - name: Complete SNB read catalogue
+        run: |
+          python3 interactive/bench/ldbc/test_harness.py
+          python3 interactive/bench/ldbc/test_snb.py
+          python3 interactive/bench/ldbc/test_updates.py
+          python3 interactive/bench/ldbc/test_operations.py
+          python3 interactive/bench/ldbc/test_cascade.py --server target/release/ddir_server
+          for workers in 1 4; do
+            python3 interactive/bench/ldbc/test_frames.py --server target/release/ddir_server \
+              --workers "$workers"
+            python3 interactive/bench/ldbc/suite.py --server target/release/ddir_server \
+              --workers "$workers" --rounds 1 --warmup 0
+          done
+          python3 interactive/bench/ldbc/suite.py --server target/release/ddir_server \
+            --mode maintained --workers 4 --rounds 1 --warmup 0
+      - name: LDBC weighted aggregates, bounded ranking and memory profiler
+        run: |
+          python3 interactive/bench/ldbc/test_aggregates.py --server target/release/ddir_server \
+            --output /tmp/ldbc-aggregates
+          python3 interactive/bench/ldbc/test_topk.py --server target/release/ddir_server \
+            --output /tmp/ldbc-topk
+          python3 interactive/bench/ldbc/test_memory_profile.py --server target/release/ddir_server \
+            --output /tmp/ldbc-memory-profile
+      - name: SNB update streams
+        run: |
+          python3 interactive/bench/ldbc/test_update_stream.py \
+            --server target/release/ddir_server --workers 4 \
+            --output /tmp/ldbc-update-stream-4
+          python3 interactive/bench/ldbc/test_update_stream.py \
+            --server target/release/ddir_server --workers 4 --check-mode replay \
+            --output /tmp/ldbc-update-stream-replay
+      - name: Authored LDBC plans agree with the generated plans
+        run: |
+          python3 interactive/bench/ldbc/authored/check.py --server target/release/ddir_server \
+            --fixture --workers 4 --reps 1 --warmup 0 --output /tmp/ldbc-authored-check
+      # Programs installed late read every import as of the time they start. A late install
+      # that reads an import at its stored times can hang (no progress for 120 s) or disagree.
+      - name: Late-install difftest sweep
+        run: |
+          DIFFTEST_SEED=40000000 DIFFTEST_CASES=500 DIFFTEST_ORACLES=split,bind,reinstall \
+            cargo test --release -p interactive --test difftest -- --ignored --nocapture
       # The explanation suite's heavier sweeps (soundness regression, fuzz)
       # are #[ignore]d for local debug runs but must gate merges. The metric
       # report is skipped (it prints, it does not assert).
       - name: Explanation suite (release, incl. ignored)
         run: cargo test --release -p interactive --test explain -- --include-ignored --skip report_demand_excess
+      # Multitemporal programs against brute-force oracles, ignored by default as they run slowly in
+      # debug: bitemporal balances and tri-temporal SCC with nested loops, at several worker counts.
+      - name: Multitemporal oracles
+        run: |
+          cargo test --release -p interactive --test bitemporal -- --include-ignored
+          cargo test --release -p interactive --test tritemporal -- --include-ignored
 
   # Check for clippy warnings
   clippy:

Cargo.toml

@@ -5,11 +5,7 @@ members = [
     "dogsdogsdogs",
     "experiments",
     "interactive",
-    "server",
-    "server/dataflows/degr_dist",
-    "server/dataflows/neighborhood",
-    "server/dataflows/random_graph",
-    "server/dataflows/reachability",
+    "interactive/server",
     #"tpchlike",
     #"doop",
     "mdbook",
@@ -19,13 +15,13 @@ resolver = "2"
 
 [workspace.package]
 edition = "2021"
-rust-version = "1.86"
+rust-version = "1.88"
 
 [workspace.dependencies]
 differential-dataflow = { path = "differential-dataflow", default-features = false, version = "0.25.1" }
-timely = { version = "0.31", default-features = false }
-columnar = { version = "0.13", default-features = false }
-#timely = { git = "https://github.com/TimelyDataflow/timely-dataflow", default-features = false }
+#timely = { version = "0.31", default-features = false }
+columnar = { version = "0.13.3", default-features = false }
+timely = { version = "0.32", default-features = false }
 #timely = { path = "../timely-dataflow/timely/", default-features = false }
 
 [workspace.lints.clippy]
@@ -97,4 +93,5 @@ debug = true
 rpath = false
 lto = true
 debug-assertions = false
-codegen-units = 4
+# One unit, so that A/B benchmarking is maximally accurate.
+codegen-units = 1