Add pure-sql comparison run to integration (#507)

jaggederest · web-flow · commit 65522ddcad50 · 2025-09-26T17:35:17.000-07:00
* Add pure-sql comparison run to integration

* rework sql to be a bit cleaner, one query per file

* working sql suite, fully specified DSNs with gssenc disabled

* CI rework

* each integration in its own run

* revert to shared suite, distributing artifacts is colossally slow

* stub GSSEnc for psycopg

* try draining and fixing out of sync errors

* more draining

* move python to restart the server before run, tweak logic change around errored connections

* revert ensure_in_sync changes

* ok I guess we dont get to keep it running
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
@@ -138,48 +138,32 @@ jobs:
           fi
           echo "Using instrumented binary at $BIN_PATH"
           echo "PGDOG_BIN=$(realpath "$BIN_PATH")" >> "$GITHUB_ENV"
-      - name: Load balancer
-        run: bash integration/load_balancer/run.sh
       - name: pgbench
         run: bash integration/pgbench/run.sh
-      - name: Verify coverage (pgbench)
-        run: bash integration/verify_profiles.sh pgbench
       - name: Go
         run: bash integration/go/run.sh
-      - name: Verify coverage (go)
-        run: bash integration/verify_profiles.sh go
       - name: JavaScript
         run: bash integration/js/pg_tests/run.sh
-      - name: Verify coverage (javascript)
-        run: bash integration/verify_profiles.sh javascript
-      - name: Toxi
-        run: bash integration/toxi/run.sh
-      - name: Verify coverage (toxi)
-        run: bash integration/verify_profiles.sh toxi
-      - name: Python
-        run: bash integration/python/run.sh
-      - name: Verify coverage (python)
-        run: bash integration/verify_profiles.sh python
       - name: Ruby
         run: bash integration/ruby/run.sh
-      - name: Verify coverage (ruby)
-        run: bash integration/verify_profiles.sh ruby
       - name: Java
         run: bash integration/java/run.sh
-      - name: Verify coverage (java)
-        run: bash integration/verify_profiles.sh java
-      - name: More complex stuff
-        run: bash integration/complex/run.sh
-      - name: Verify coverage (complex)
-        run: bash integration/verify_profiles.sh complex
+      - name: SQL
+        run: bash integration/sql/run.sh
+      - name: Toxi
+        run: bash integration/toxi/run.sh
       - name: Rust
         run: bash integration/rust/run.sh
-      - name: Verify coverage (rust)
-        run: bash integration/verify_profiles.sh rust
+      - name: Stop shared PgDog
+        run: bash -lc 'source integration/common.sh; stop_pgdog'
+      - name: Python
+        run: bash integration/python/run.sh
+      - name: Load balancer
+        run: bash integration/load_balancer/run.sh
+      - name: More complex stuff
+        run: bash integration/complex/run.sh
       - name: Dry run
         run: bash integration/dry_run/run.sh
-      - name: Verify coverage (dry_run)
-        run: bash integration/verify_profiles.sh dry_run
       # - name: Plugins
       #   run: bash integration/plugins/run.sh
       - name: Ensure PgDog stopped
diff --git a/.gitignore b/.gitignore
@@ -53,3 +53,4 @@ CLAUDE.local.md
 pgdog-plugin/src/bindings.rs
 local/
 integration/log.txt
+.pycache
diff --git a/integration/common.sh b/integration/common.sh
@@ -54,6 +54,9 @@ function run_pgdog() {
 }
 
 function stop_pgdog() {
+    if [ "${PGDOG_KEEP_RUNNING:-0}" = "1" ]; then
+        return
+    fi
     local pid_file="${COMMON_DIR}/pgdog.pid"
     local config_file="${COMMON_DIR}/pgdog.config"
     if [ -f "${pid_file}" ]; then
diff --git a/integration/load_balancer/run.sh b/integration/load_balancer/run.sh
@@ -37,7 +37,5 @@ popd
 
 stop_pgdog
 
-PGDOG_NO_RESTART=1 bash ${SCRIPT_DIR}/../verify_profiles.sh load_balancer
-
 docker-compose down
 popd
diff --git a/integration/pgdog.toml b/integration/pgdog.toml
@@ -218,6 +218,33 @@ name = "sharded_uuid"
 column = "id_uuid"
 data_type = "uuid"
 
+# ------------------------------------------------------------------------------
+# ----- SQL Regression Sample --------------------------------------------------
+
+[[sharded_tables]]
+database = "pgdog_sharded"
+name = "sql_regression_sample"
+column = "id"
+data_type = "bigint"
+
+[[sharded_mappings]]
+database = "pgdog_sharded"
+table = "sql_regression_sample"
+column = "id"
+kind = "range"
+start = 0
+end = 100
+shard = 0
+
+[[sharded_mappings]]
+database = "pgdog_sharded"
+table = "sql_regression_sample"
+column = "id"
+kind = "range"
+start = 100
+end = 200
+shard = 1
+
 # ------------------------------------------------------------------------------
 # ----- Range Sharded :: BIGINT ------------------------------------------------
 
diff --git a/integration/python/run.sh b/integration/python/run.sh
@@ -1,5 +1,5 @@
 #!/bin/bash
-set -e
+set -euo pipefail
 SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )
 source ${SCRIPT_DIR}/../common.sh
 
diff --git a/integration/run.sh b/integration/run.sh
@@ -5,4 +5,5 @@ pushd ${SCRIPT_DIR}
 bash python/run.sh
 bash ruby/run.sh
 bash java/run.sh
+bash sql/run.sh
 popd
diff --git a/integration/sql/README.md b/integration/sql/README.md
@@ -0,0 +1,41 @@
+# SQL Regression Suite
+
+This directory hosts the pure-SQL regression harness used to ensure PgDog matches the behaviour of direct PostgreSQL connections.
+
+## Layout
+
+- `cases/` – numbered SQL scenarios using the `00x_slug_(setup|case|teardown).sql` convention.
+- `global_setup.sql` / `global_teardown.sql` – optional hooks that run before/after every case.
+- `lib.py` – harness logic plus in-code definitions for the six comparison targets (DSNs disable GSS encryption so PgDog can proxy them).
+- `run.sh` / `dev.sh` – integration entrypoints matching the other language suites.
+- `test_sql_regression.py` – pytest runner that executes every case against all registered targets.
+
+## Workflow
+
+1. `run.sh` builds PgDog, starts the proxy, runs the suite, then shuts PgDog down.
+2. For each scenario the harness:
+   - runs `global_setup.sql` (if present) and the scenario’s `*_setup.sql` through every distinct DSN used by the targets (shared once for baseline/pgdog, once for the sharded pool),
+   - executes the `*_case.sql` statements against all six variants (baseline Postgres, PgDog, PgDog sharded × text/binary),
+   - asserts that statement status, column names, column types, row counts, and row payloads are identical for every target,
+   - runs the scenario’s `*_teardown.sql` and then `global_teardown.sql` (if present) across the same DSNs to leave databases clean.
+
+Add more scenarios by dropping files into `cases/`:
+
+```
+cases/
+  002_select_edge_case_setup.sql    # optional
+  002_select_edge_case_case.sql     # required
+  002_select_edge_case_teardown.sql # optional
+```
+
+Metadata lives in SQL comments at the top of the `*_case.sql` file:
+
+```
+-- description: Exercise numeric encoding edge cases
+-- tags: standard sharded
+-- transactional: true
+-- skip-targets: pgdog_sharded_binary
+-- only-targets: postgres_standard_text pgdog_standard_text
+```
+
+Tags are informative only. Use `skip-targets` to drop one or more target names, or `only-targets` to pin a case to a specific subset (order-sensitive). Set `transactional: false` when the statements must commit.
diff --git a/integration/sql/__init__.py b/integration/sql/__init__.py
diff --git a/integration/sql/cases/001_sample_case.sql b/integration/sql/cases/001_sample_case.sql
@@ -0,0 +1,5 @@
+-- description: Simple smoke test ensuring identical behaviour for constants and table data
+-- tags: standard
+-- transactional: true
+
+SELECT id, val, created_at FROM sql_regression_sample ORDER BY id;
diff --git a/integration/sql/cases/001_sample_setup.sql b/integration/sql/cases/001_sample_setup.sql
@@ -0,0 +1,10 @@
+DROP TABLE IF EXISTS sql_regression_sample;
+CREATE TABLE sql_regression_sample (
+    id BIGINT PRIMARY KEY,
+    val TEXT,
+    created_at TIMESTAMPTZ
+);
+-- Insert rows individually so PgDog can route each id to a single shard.
+INSERT INTO sql_regression_sample (id, val, created_at) VALUES (1, 'alpha', TIMESTAMPTZ '2024-01-01T00:00:00Z');
+INSERT INTO sql_regression_sample (id, val, created_at) VALUES (2, 'beta', TIMESTAMPTZ '2024-01-01T00:00:00Z');
+INSERT INTO sql_regression_sample (id, val, created_at) VALUES (3, 'gamma', TIMESTAMPTZ '2024-01-01T00:00:00Z');
diff --git a/integration/sql/cases/001_sample_teardown.sql b/integration/sql/cases/001_sample_teardown.sql
@@ -0,0 +1 @@
+DROP TABLE IF EXISTS sql_regression_sample;
diff --git a/integration/sql/dev.sh b/integration/sql/dev.sh
@@ -0,0 +1,16 @@
+#!/bin/bash
+set -euo pipefail
+SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )
+
+pushd ${SCRIPT_DIR} >/dev/null
+
+if [[ ! -d venv ]]; then
+    python3 -m venv venv
+fi
+source venv/bin/activate
+pip install --upgrade pip >/dev/null
+pip install -r requirements.txt
+
+pytest -x
+
+popd >/dev/null
diff --git a/integration/sql/global_setup.sql b/integration/sql/global_setup.sql
@@ -0,0 +1,2 @@
+-- Global setup executed before every SQL regression scenario.
+-- Add shared fixture prep (e.g., CREATE SCHEMA) as needed.
diff --git a/integration/sql/global_teardown.sql b/integration/sql/global_teardown.sql
@@ -0,0 +1,2 @@
+-- Global teardown executed after every SQL regression scenario.
+-- Use this file to drop shared fixtures created in global_setup.sql.
diff --git a/integration/sql/lib.py b/integration/sql/lib.py
diff --git a/integration/sql/requirements.txt b/integration/sql/requirements.txt
diff --git a/integration/sql/run.sh b/integration/sql/run.sh
diff --git a/integration/sql/test_sql_regression.py b/integration/sql/test_sql_regression.py
diff --git a/pgdog/src/frontend/listener.rs b/pgdog/src/frontend/listener.rs
diff --git a/pgdog/src/net/messages/hello.rs b/pgdog/src/net/messages/hello.rs

Original file line number	Diff line number	Diff line change
`@@ -54,6 +54,9 @@ function run_pgdog() {`
`54`	`54`	`}`
`55`	`55`
`56`	`56`	`function stop_pgdog() {`
	`57`	`+ if [ "${PGDOG_KEEP_RUNNING:-0}" = "1" ]; then`
	`58`	`+ return`
	`59`	`+ fi`
`57`	`60`	`local pid_file="${COMMON_DIR}/pgdog.pid"`
`58`	`61`	`local config_file="${COMMON_DIR}/pgdog.config"`
`59`	`62`	`if [ -f "${pid_file}" ]; then`