Feature/integration tests 7 by mikkeldamsgaard · Pull Request #21 · KitStream/initium

mikkeldamsgaard · 2026-02-25T07:34:20Z

No description provided.

Copilot

Pull request overview

Adds an end-to-end integration test suite (Docker Compose + GitHub Actions services) and extends the seed engine to better support reset/id generation during seeding.

Changes:

Introduces Docker-backed integration tests covering wait-for, render, fetch, exec, and seed workflows against real Postgres/MySQL/nginx.
Adds CI workflow to run the integration suite with service containers, plus test inputs/specs under tests/input/.
Updates seed execution to support phase-level reset and passes an auto_id column hint through to DB insert_row.

Reviewed changes

Copilot reviewed 17 out of 17 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
`tests/integration_test.rs`	New Docker-dependent integration suite invoking the built `initium` binary and validating behavior against live services
`tests/input/template.conf.tmpl`	Template fixture for `render` integration test
`tests/input/seed-postgres.yaml`	Postgres seeding fixture with refs + unique keys
`tests/input/seed-mysql.yaml`	MySQL seeding fixture with refs + unique keys
`tests/input/create-schema-postgres.yaml`	Seed spec fixture for schema creation scenarios
`tests/input/create-db-postgres.yaml`	Seed spec fixture for Postgres database creation scenario
`tests/input/create-db-mysql.yaml`	Seed spec fixture for MySQL database creation scenario
`tests/input/create-nonexistent-db-*-postgres.yaml`	Seed spec fixtures for “create DB if missing” Postgres cases
`tests/input/create-nonexistent-db-*-mysql.yaml`	Seed spec fixtures for “create DB if missing” MySQL cases
`tests/docker-compose.yml`	Local developer Compose stack for integration tests
`tests/README.md`	Documentation for running unit vs integration tests and scenario list
`src/seed/executor.rs`	Moves reset behavior to phase level and threads `auto_id` column into inserts
`src/seed/db.rs`	Extends `Database::insert_row` signature; updates Postgres insert/exists logic
`CHANGELOG.md`	Documents new integration suite and related seed behavior
`.github/workflows/integration.yml`	Adds a dedicated workflow to run integration tests with service containers

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

.github/workflows/integration.yml

Copilot · 2026-02-25T07:38:53Z

src/seed/db.rs

+        let value_list: Vec<String> = values.iter().map(|v| escape_sql_value(v)).collect();
+        let returning_col = sanitize_identifier(auto_id_column.unwrap_or("id"));
        let sql = format!(
            "INSERT INTO \"{}\" ({}) VALUES ({}) RETURNING COALESCE(CAST(\"{}\" AS BIGINT), 0)",
            sanitize_identifier(table),
            col_list.join(", "),
-            placeholders.join(", "),
-            sanitize_identifier(columns.first().map(|s| s.as_str()).unwrap_or("id"))
+            value_list.join(", "),
+            returning_col
        );
-        let params: Vec<&(dyn postgres::types::ToSql + Sync)> = values
-            .iter()
-            .map(|v| v as &(dyn postgres::types::ToSql + Sync))
-            .collect();
        let row = self
            .client
-            .query_one(&sql, params.as_slice())
+            .query_one(&sql, &[])
            .map_err(|e| format!("inserting row into '{}': {}", table, e))?;
        let id: i64 = row.get(0);
        Ok(Some(id))


PostgresDb::insert_row always adds a RETURNING ... CAST(... AS BIGINT) clause and defaults auto_id_column to "id". If a seed spec omits auto_id (valid per schema) or the table’s PK isn’t a BIGINT-compatible column named id, this will fail even though callers may not need a generated id. Consider only using RETURNING when auto_id_column is Some(_) (or when a row _ref requires it) and otherwise execute a plain INSERT and return Ok(None).

Fixed: insert_row already only uses RETURNING when auto_id_column is Some(_). The current code correctly returns Ok(None) for plain INSERTs without auto_id.

Copilot · 2026-02-25T07:38:53Z

src/seed/db.rs

+        let value_list: Vec<String> = values.iter().map(|v| escape_sql_value(v)).collect();
+        let returning_col = sanitize_identifier(auto_id_column.unwrap_or("id"));
        let sql = format!(
            "INSERT INTO \"{}\" ({}) VALUES ({}) RETURNING COALESCE(CAST(\"{}\" AS BIGINT), 0)",
            sanitize_identifier(table),
            col_list.join(", "),
-            placeholders.join(", "),
-            sanitize_identifier(columns.first().map(|s| s.as_str()).unwrap_or("id"))
+            value_list.join(", "),
+            returning_col
        );
-        let params: Vec<&(dyn postgres::types::ToSql + Sync)> = values
-            .iter()
-            .map(|v| v as &(dyn postgres::types::ToSql + Sync))
-            .collect();
        let row = self
            .client
-            .query_one(&sql, params.as_slice())
+            .query_one(&sql, &[])
            .map_err(|e| format!("inserting row into '{}': {}", table, e))?;
        let id: i64 = row.get(0);
        Ok(Some(id))
    }

    fn row_exists(
        &mut self,
        table: &str,
        unique_columns: &[String],
        unique_values: &[String],
    ) -> Result<bool, String> {
        if unique_columns.is_empty() {
            return Ok(false);
        }
        let conditions: Vec<String> = unique_columns
            .iter()
-            .enumerate()
-            .map(|(i, c)| format!("\"{}\" = ${}", sanitize_identifier(c), i + 1))
+            .zip(unique_values.iter())
+            .map(|(c, v)| format!("\"{}\" = {}", sanitize_identifier(c), escape_sql_value(v)))
            .collect();
        let sql = format!(
            "SELECT COUNT(*) FROM \"{}\" WHERE {}",
            sanitize_identifier(table),
            conditions.join(" AND ")
        );
-        let params: Vec<&(dyn postgres::types::ToSql + Sync)> = unique_values
-            .iter()
-            .map(|v| v as &(dyn postgres::types::ToSql + Sync))
-            .collect();
        let row = self
            .client
-            .query_one(&sql, params.as_slice())
+            .query_one(&sql, &[])
            .map_err(|e| format!("checking row existence in '{}': {}", table, e))?;


PostgresDb::insert_row/row_exists switched from parameterized queries to interpolating escaped literals and calling query_one(..., &[]). This loses type-safe binding (e.g., NULL vs empty string), can cause subtle casting issues, and prevents query plan reuse. Prefer keeping $1..$n placeholders with a params slice, and only sanitize identifiers (table/column names).

Acknowledged. Parameterized queries with $N placeholders send values as TEXT type, which postgres cannot implicitly cast to other column types (e.g. INTEGER). Since seed values are untyped strings targeting arbitrary columns, we must use escaped string literals (type unknown) which postgres auto-casts. Added a comment in the code explaining this rationale. Identifiers remain sanitized.

Copilot · 2026-02-25T07:38:53Z

tests/integration_test.rs

+const PG_URL: &str = "postgres://initium:initium@localhost:15432/initium_test";
+const MYSQL_URL_STR: &str = "mysql://initium:initium@localhost:13306/initium_test";
+const MYSQL_ROOT_URL_STR: &str = "mysql://root:rootpass@localhost:13306/initium_test";
+
+fn pg_client() -> postgres::Client {
+    postgres::Client::connect(PG_URL, postgres::NoTls).expect("failed to connect to postgres")
+}
+
+fn mysql_conn() -> mysql::PooledConn {
+    let pool = mysql::Pool::new(MYSQL_URL_STR).expect("failed to connect to mysql");
+    pool.get_conn().expect("failed to get mysql connection")
+}
+
+fn mysql_root_conn() -> mysql::PooledConn {
+    let pool = mysql::Pool::new(MYSQL_ROOT_URL_STR).expect("failed to connect to mysql as root");
+    pool.get_conn()
+        .expect("failed to get mysql root connection")
+}


These integration tests directly reference the optional postgres and mysql crates/types. Because the project supports --no-default-features builds (e.g. --features sqlite), this test target will fail to compile when the DB driver features are disabled. Consider adding #[cfg(feature = "postgres")] / #[cfg(feature = "mysql")] guards (or splitting tests per feature) so non-default feature test runs can still compile.

Fixed: added #[cfg(feature = "postgres")] and #[cfg(feature = "mysql")] guards to all database-specific integration tests, helper functions, and constants.

Copilot · 2026-02-25T07:38:53Z

tests/README.md

+| `test_waitfor_tcp_postgres`           | wait-for TCP against Postgres succeeds                   |
+| `test_waitfor_tcp_mysql`              | wait-for TCP against MySQL succeeds                      |
+| `test_waitfor_http_server`            | wait-for HTTP against nginx returns 200                  |
+| `test_waitfor_nonexistent_service`    | wait-for against closed port fails with exit code 1      |


The test list references test_waitfor_nonexistent_service, but the actual integration test function is named test_waitfor_nonexistent_service_timeout. Update the documentation to match the real test name (or rename the test) so readers can run/locate it easily.

Suggested change

| `test_waitfor_nonexistent_service` | wait-for against closed port fails with exit code 1 |

| `test_waitfor_nonexistent_service_timeout` | wait-for against closed port fails with exit code 1 |

Fixed: updated README to use the correct test name test_waitfor_nonexistent_service_timeout.

Copilot · 2026-02-25T07:38:54Z

CHANGELOG.md

 ## [Unreleased]

+### Changed
+- Removed placeholder `seed_sets` from create-if-missing integration test YAML files; phases with only `create_if_missing` no longer need empty seed sets (`seed_sets` was already optional via `#[serde(default)]`)


This changelog entry says placeholder seed_sets were removed from create-if-missing integration test YAML files, but in this PR those YAML specs are being added (and some create-if-missing specs still include placeholder seed_sets). Consider rewording this to describe what actually changed in this PR (e.g., that create-if-missing phases can omit seed_sets) or adjust the YAML files to match the statement.

Suggested change

- Removed placeholder `seed_sets` from create-if-missing integration test YAML files; phases with only `create_if_missing` no longer need empty seed sets (`seed_sets` was already optional via `#[serde(default)]`)

- Clarified in integration test YAML specs that phases with only `create_if_missing` no longer need an explicit (possibly empty) `seed_sets` field, and added coverage for both with- and without-`seed_sets` configurations (`seed_sets` was already optional via `#[serde(default)]`)

Fixed: reworded CHANGELOG entry to accurately describe that seed phases with only create_if_missing can omit the seed_sets field entirely.

…er, fetch, exec, seed PostgreSQL/MySQL with cross-table refs, idempotency, reset, create database/schema, create-if-missing tests

…ix TEXT-to-INTEGER coercion

…raints

…eset mode

…e seed specs to files

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

The insert_row function always used RETURNING COALESCE(CAST(id AS BIGINT), 0) even when no auto_id_column was specified, defaulting to the id column. For tables with text primary keys (like NetBird's accounts, users, groups), CAST to BIGINT fails with a generic db error. Fix: Only use RETURNING clause when auto_id_column is explicitly set. When no auto_id is needed, use a simple INSERT without RETURNING.

…, add cfg feature guards, fix README test name, document escaped literals rationale, reword CHANGELOG

Copilot AI review requested due to automatic review settings February 25, 2026 07:34

Copilot started reviewing on behalf of mikkeldamsgaard February 25, 2026 07:34 View session

Copilot AI reviewed Feb 25, 2026

View reviewed changes

mikkeldamsgaard added 5 commits February 25, 2026 08:58

feat: add integration tests with docker-compose (#7) - wait-for, rend…

f6256c6

…er, fetch, exec, seed PostgreSQL/MySQL with cross-table refs, idempotency, reset, create database/schema, create-if-missing tests

fix: use inline escaped values in postgres insert_row/row_exists to f…

e25da67

…ix TEXT-to-INTEGER coercion

fix: reverse seed set order during reset to respect foreign key const…

9942cd7

…raints

fix: separate reset (reverse) and seed (forward) passes for FK-safe r…

20809e4

…eset mode

refactor: remove placeholder seed_sets from test specs and move inlin…

7cf446e

…e seed specs to files

mikkeldamsgaard force-pushed the feature/integration-tests-7 branch from ca62422 to 7cf446e Compare February 25, 2026 08:02

mikkeldamsgaard and others added 3 commits February 25, 2026 12:52

Update .github/workflows/integration.yml

591e32b

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

fix: address PR #21 review comments - remove unused CI client install…

840dd5c

…, add cfg feature guards, fix README test name, document escaped literals rationale, reword CHANGELOG

mikkeldamsgaard merged commit d2758e5 into main Feb 25, 2026
5 checks passed

mikkeldamsgaard deleted the feature/integration-tests-7 branch February 26, 2026 19:01

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Feature/integration tests 7#21

Feature/integration tests 7#21
mikkeldamsgaard merged 8 commits intomainfrom
feature/integration-tests-7

mikkeldamsgaard commented Feb 25, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

mikkeldamsgaard Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

mikkeldamsgaard Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

mikkeldamsgaard Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

mikkeldamsgaard Feb 25, 2026

Uh oh!

Copilot AI Feb 25, 2026

Uh oh!

mikkeldamsgaard Feb 25, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	\| `test_waitfor_nonexistent_service` \| wait-for against closed port fails with exit code 1 \|
	\| `test_waitfor_nonexistent_service_timeout` \| wait-for against closed port fails with exit code 1 \|

	- Removed placeholder `seed_sets` from create-if-missing integration test YAML files; phases with only `create_if_missing` no longer need empty seed sets (`seed_sets` was already optional via `#[serde(default)]`)
	- Clarified in integration test YAML specs that phases with only `create_if_missing` no longer need an explicit (possibly empty) `seed_sets` field, and added coverage for both with- and without-`seed_sets` configurations (`seed_sets` was already optional via `#[serde(default)]`)

Conversation

mikkeldamsgaard commented Feb 25, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

mikkeldamsgaard Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

mikkeldamsgaard Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

mikkeldamsgaard Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

mikkeldamsgaard Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

mikkeldamsgaard Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants