test: integration, compile all dialects and diff (#5344)

Author: kgutwin

.config/nextest.toml

@@ -3,6 +3,11 @@ fail-fast = false
 failure-output = "final"
 slow-timeout = {period = "500ms", terminate-after = 4}
 
+[[profile.default.overrides]]
+# compileall does many passes over the same query, so it takes a bit longer
+filter = 'package(prqlc) & test(queries::compileall::)'
+slow-timeout = {period = "10s", terminate-after = 2}
+
 [[profile.default.overrides]]
 filter = 'package(prqlc) & test(queries::results::)'
 slow-timeout = {period = "10s", terminate-after = 4}

prqlc/prqlc/tests/integration/queries.rs

@@ -5,6 +5,7 @@ use std::{env, fs};
 use insta::assert_debug_snapshot;
 use insta::{assert_snapshot, with_settings};
 use prqlc::sql::Dialect;
+use prqlc::sql::SupportLevel;
 use prqlc::{Options, Target};
 use test_each_file::test_each_path;
 
@@ -50,6 +51,64 @@ mod compile {
     }
 }
 
+fn should_run_query(dialect: Dialect, prql: &str) -> bool {
+    let dialect_str = dialect.to_string().to_lowercase();
+
+    match dialect.support_level() {
+        SupportLevel::Supported => !prql.contains(&format!("{dialect_str}:skip")),
+        SupportLevel::Unsupported => prql.contains(&format!("{dialect_str}:test")),
+        SupportLevel::Nascent => false,
+    }
+}
+
+mod compileall {
+    use super::*;
+    use similar::TextDiff;
+    use strum::IntoEnumIterator;
+
+    test_each_path! { in "./prqlc/prqlc/tests/integration/queries" => run }
+
+    fn run(prql_path: &Path) {
+        let test_name = prql_path.file_stem().unwrap().to_str().unwrap();
+        let prql = fs::read_to_string(prql_path).unwrap();
+        if prql.contains("generic:skip") {
+            return;
+        }
+
+        // first compile with the generic dialect
+        let target = Target::Sql(Some(Dialect::Generic));
+        let options = Options::default().no_signature().with_target(target);
+
+        let generic_sql = prqlc::compile(&prql, &options).unwrap();
+
+        // next compile with each dialect
+        let mut diffsnap = "".to_owned();
+        for dialect in Dialect::iter() {
+            if !should_run_query(dialect, &prql) {
+                continue;
+            }
+
+            let dialect_target = Target::Sql(Some(dialect));
+            let dialect_options = Options::default()
+                .no_signature()
+                .with_target(dialect_target);
+
+            let dialect_sql = prqlc::compile(&prql, &dialect_options).unwrap();
+
+            let diff = TextDiff::from_lines(&generic_sql, &dialect_sql);
+            diffsnap = format!(
+                "{diffsnap}\n{}",
+                diff.unified_diff()
+                    .context_radius(10)
+                    .header("generic", &dialect.to_string())
+            );
+        }
+        with_settings!({ input_file => prql_path }, {
+            assert_snapshot!(test_name, diffsnap, &prql)
+        });
+    }
+}
+
 mod fmt {
     use super::*;
 
@@ -96,23 +155,12 @@ mod debug_lineage {
 mod results {
 
     use itertools::Itertools;
-    use prqlc::sql::SupportLevel;
 
     use super::*;
     use crate::dbs::{batch_to_csv, runners};
 
     test_each_path! { in "./prqlc/prqlc/tests/integration/queries" => run }
 
-    fn should_run_query(dialect: Dialect, prql: &str) -> bool {
-        let dialect_str = dialect.to_string().to_lowercase();
-
-        match dialect.support_level() {
-            SupportLevel::Supported => !prql.contains(&format!("{dialect_str}:skip")),
-            SupportLevel::Unsupported => prql.contains(&format!("{dialect_str}:test")),
-            SupportLevel::Nascent => false,
-        }
-    }
-
     fn run(prql_path: &Path) {
         let test_name = prql_path.file_stem().unwrap().to_str().unwrap();
         let prql = fs::read_to_string(prql_path).unwrap();

prqlc/prqlc/tests/integration/snapshots/integration__queries__compileall__aggregation.snap

@@ -0,0 +1,20 @@
+---
+source: prqlc/prqlc/tests/integration/queries.rs
+expression: "# mysql:skip\n# clickhouse:skip\n# glaredb:skip (the string_agg function is not supported)\nfrom tracks\nfilter genre_id == 100\nderive empty_name = name == ''\naggregate {sum track_id, concat_array name, all empty_name, any empty_name}\n"
+input_file: prqlc/prqlc/tests/integration/queries/aggregation.prql
+---
+--- generic
++++ sqlite
+@@ -1,9 +1,9 @@
+ SELECT
+   COALESCE(SUM(track_id), 0),
+-  COALESCE(STRING_AGG(name, ''), ''),
+-  COALESCE(BOOL_AND(name = ''), TRUE),
+-  COALESCE(BOOL_OR(name = ''), FALSE)
++  COALESCE(GROUP_CONCAT(name, ''), ''),
++  COALESCE(MIN(name = '') > 0, TRUE),
++  COALESCE(MAX(name = '') > 0, FALSE)
+ FROM
+   tracks
+ WHERE
+   genre_id = 100

prqlc/prqlc/tests/integration/snapshots/integration__queries__compileall__append_select.snap

@@ -0,0 +1,6 @@
+---
+source: prqlc/prqlc/tests/integration/queries.rs
+expression: "from invoices\nselect { customer_id, invoice_id, billing_country }\ntake 10..15\nappend (\n  from invoices\n  select { customer_id, invoice_id, billing_country }\n  take 40..45\n)\nselect { billing_country, invoice_id }\n"
+input_file: prqlc/prqlc/tests/integration/queries/append_select.prql
+---
+

prqlc/prqlc/tests/integration/snapshots/integration__queries__compileall__append_select_compute.snap

@@ -0,0 +1,41 @@
+---
+source: prqlc/prqlc/tests/integration/queries.rs
+expression: "from invoices\nselect { customer_id, invoice_id, total }\ntake 5\nappend (\n  from invoice_items\n  select { invoice_line_id, invoice_id, unit_price }\n  take 5\n)\nselect { a = customer_id * 2, b = math.round 1 (invoice_id * total) }\n"
+input_file: prqlc/prqlc/tests/integration/queries/append_select_compute.prql
+---
+--- generic
++++ glaredb
+@@ -23,13 +23,13 @@
+     ) AS table_2
+   UNION
+   ALL
+   SELECT
+     *
+   FROM
+     table_0
+ )
+ SELECT
+   customer_id * 2 AS a,
+-  ROUND(invoice_id * total, 1) AS b
++  ROUND((invoice_id * total)::numeric, 1) AS b
+ FROM
+   table_1
+
+
+--- generic
++++ postgres
+@@ -23,13 +23,13 @@
+     ) AS table_2
+   UNION
+   ALL
+   SELECT
+     *
+   FROM
+     table_0
+ )
+ SELECT
+   customer_id * 2 AS a,
+-  ROUND(invoice_id * total, 1) AS b
++  ROUND((invoice_id * total)::numeric, 1) AS b
+ FROM
+   table_1

prqlc/prqlc/tests/integration/snapshots/integration__queries__compileall__append_select_multiple_with_null.snap

@@ -0,0 +1,6 @@
+---
+source: prqlc/prqlc/tests/integration/queries.rs
+expression: "from invoices\nselect { customer_id, invoice_id, billing_country }\ntake 5\nappend (\n  from employees\n  select { employee_id, employee_id, country }\n  take 5\n)\nappend (\n  from invoice_items\n  select { invoice_line_id, invoice_id, null }\n  take 5\n)\nselect { billing_country, invoice_id }\n"
+input_file: prqlc/prqlc/tests/integration/queries/append_select_multiple_with_null.prql
+---
+