feat(ir): more flexible dereferencing support for join right hand side (ibis-project#8992)

Enables to use fields from parent tables of the join right hand side
instead of enforcing to use the same exact table:

```py
t1 = ibis.table(name="t1", schema={"a": "int64", "b": "string"})
t2 = ibis.table(name="t2", schema={"c": "int64", "d": "string"})

t3 = t2.mutate(e=t2.c + 1)
joined = t1.join(t3, [t1.a == t2.c])  # here we use t2.c instead of t3.c
```

Identify ambiguous cases and raise an error, like the following case:

```py
t.join(t, [t.a == t.a])
```

depends on:
- ibis-project#9043 
- ibis-project#9041 

fixes ibis-project#8581

Author: kszucs

docs/_code/setup_penguins.qmd

@@ -1,6 +1,7 @@
 ```{python}
 import ibis  # <1>
 import ibis.selectors as s  # <1>
+from ibis import _
 
 ibis.options.interactive = True  # <2>
 

docs/how-to/analytics/basics.qmd

@@ -58,20 +58,20 @@ t.mutate(bill_length_cm=t["bill_length_mm"] / 10).relocate(
 Use the `.join()` method to join data:
 
 ```{python}
-t.join(t, t["species"] == t["species"], how="left_semi")
+t.join(t, ["species"], how="left_semi")
 ```
 
 ## Combining it all together
 
 We can use [the underscore to chain expressions together](./chain_expressions.qmd).
 
 ```{python}
-t.join(t, t["species"] == t["species"], how="left_semi").filter(
-    ibis._["species"] != "Adelie"
+t.join(t, ["species"], how="left_semi").filter(
+    _.species != "Adelie"
 ).group_by(["species", "island"]).aggregate(
-    avg_bill_length=ibis._["bill_length_mm"].mean()
+    avg_bill_length=_.bill_length_mm.mean()
 ).order_by(
-    ibis._["avg_bill_length"].desc()
+    _.avg_bill_length.desc()
 )
 ```
 

ibis/backends/bigquery/tests/unit/snapshots/test_compiler/test_projection_fusion_only_peeks_at_immediate_parent/out.sql

@@ -15,5 +15,5 @@ SELECT
   `t3`.`val`,
   `t3`.`XYZ`
 FROM `t1` AS `t3`
-INNER JOIN `t1` AS `t5`
+INNER JOIN `t1` AS `t4`
   ON TRUE

ibis/backends/bigquery/tests/unit/test_compiler.py

@@ -274,7 +274,8 @@ class MockBackend(ibis.backends.bigquery.Backend):
     table = ops.SQLQueryResult("select * from t", schema, ibis_client).to_expr()
     for _ in range(num_joins):  # noqa: F402
         table = table.mutate(dummy=ibis.literal(""))
-        table = table.left_join(table, ["dummy"])[[table]]
+        table_ = table.view()
+        table = table.left_join(table_, ["dummy"])[[table_]]
 
     start = time.time()
     table.compile()

ibis/backends/clickhouse/tests/snapshots/test_select/test_join_self_reference/out.sql

@@ -13,5 +13,5 @@ SELECT
   "t1"."year",
   "t1"."month"
 FROM "functional_alltypes" AS "t1"
-INNER JOIN "functional_alltypes" AS "t3"
-  ON "t1"."id" = "t3"."id"
+INNER JOIN "functional_alltypes" AS "t2"
+  ON "t1"."id" = "t2"."id"

ibis/backends/impala/tests/snapshots/test_sql/test_join_key_name2/out.sql

@@ -1 +1 @@
-WITH `t9` AS (SELECT EXTRACT(year FROM `t8`.`odate`) AS `year`, COUNT(*) AS `CountStar()` FROM (SELECT `t6`.`c_custkey`, `t6`.`c_name`, `t6`.`c_address`, `t6`.`c_nationkey`, `t6`.`c_phone`, `t6`.`c_acctbal`, `t6`.`c_mktsegment`, `t6`.`c_comment`, `t4`.`r_name` AS `region`, `t7`.`o_totalprice`, CAST(`t7`.`o_orderdate` AS TIMESTAMP) AS `odate` FROM `tpch_region` AS `t4` INNER JOIN `tpch_nation` AS `t5` ON `t4`.`r_regionkey` = `t5`.`n_regionkey` INNER JOIN `tpch_customer` AS `t6` ON `t6`.`c_nationkey` = `t5`.`n_nationkey` INNER JOIN `tpch_orders` AS `t7` ON `t7`.`o_custkey` = `t6`.`c_custkey`) AS `t8` GROUP BY 1) SELECT `t11`.`year`, `t11`.`CountStar()` AS `pre_count`, `t13`.`CountStar()` AS `post_count` FROM `t9` AS `t11` INNER JOIN `t9` AS `t13` ON `t11`.`year` = `t13`.`year`
+WITH `t9` AS (SELECT EXTRACT(year FROM `t8`.`odate`) AS `year`, COUNT(*) AS `CountStar()` FROM (SELECT `t6`.`c_custkey`, `t6`.`c_name`, `t6`.`c_address`, `t6`.`c_nationkey`, `t6`.`c_phone`, `t6`.`c_acctbal`, `t6`.`c_mktsegment`, `t6`.`c_comment`, `t4`.`r_name` AS `region`, `t7`.`o_totalprice`, CAST(`t7`.`o_orderdate` AS TIMESTAMP) AS `odate` FROM `tpch_region` AS `t4` INNER JOIN `tpch_nation` AS `t5` ON `t4`.`r_regionkey` = `t5`.`n_regionkey` INNER JOIN `tpch_customer` AS `t6` ON `t6`.`c_nationkey` = `t5`.`n_nationkey` INNER JOIN `tpch_orders` AS `t7` ON `t7`.`o_custkey` = `t6`.`c_custkey`) AS `t8` GROUP BY 1) SELECT `t11`.`year`, `t11`.`CountStar()` AS `pre_count`, `t12`.`CountStar()` AS `post_count` FROM `t9` AS `t11` INNER JOIN `t9` AS `t12` ON `t11`.`year` = `t12`.`year`

ibis/backends/impala/tests/snapshots/test_sql/test_join_with_nested_or_condition/out.sql

@@ -2,12 +2,12 @@ SELECT
   `t1`.`a`,
   `t1`.`b`
 FROM `t` AS `t1`
-INNER JOIN `t` AS `t3`
-  ON `t1`.`a` = `t3`.`a`
+INNER JOIN `t` AS `t2`
+  ON `t1`.`a` = `t2`.`a`
   AND (
     (
-      `t1`.`a` <> `t3`.`b`
+      `t1`.`a` <> `t2`.`b`
     ) OR (
-      `t1`.`b` <> `t3`.`a`
+      `t1`.`b` <> `t2`.`a`
     )
   )

ibis/backends/impala/tests/snapshots/test_sql/test_join_with_nested_xor_condition/out.sql

@@ -2,13 +2,13 @@ SELECT
   `t1`.`a`,
   `t1`.`b`
 FROM `t` AS `t1`
-INNER JOIN `t` AS `t3`
-  ON `t1`.`a` = `t3`.`a`
+INNER JOIN `t` AS `t2`
+  ON `t1`.`a` = `t2`.`a`
   AND (
     (
-      `t1`.`a` <> `t3`.`b` OR `t1`.`b` <> `t3`.`a`
+      `t1`.`a` <> `t2`.`b` OR `t1`.`b` <> `t2`.`a`
     )
     AND NOT (
-      `t1`.`a` <> `t3`.`b` AND `t1`.`b` <> `t3`.`a`
+      `t1`.`a` <> `t2`.`b` AND `t1`.`b` <> `t2`.`a`
     )
   )

ibis/backends/impala/tests/snapshots/test_sql/test_limit_cte_extract/out.sql

@@ -19,5 +19,5 @@ SELECT
   `t3`.`year`,
   `t3`.`month`
 FROM `t1` AS `t3`
-INNER JOIN `t1` AS `t5`
+INNER JOIN `t1` AS `t4`
   ON TRUE

ibis/backends/impala/tests/snapshots/test_sql/test_nested_join_base/out.sql

@@ -7,7 +7,7 @@ WITH `t1` AS (
     1
 )
 SELECT
-  `t5`.`uuid`,
+  `t3`.`uuid`,
   `t3`.`CountStar(t)`
 FROM (
   SELECT