amp: Change naming convention for CTEs

The old convention made queries very hard to read. We now simply demand
that users stay away from tables etc. whose name starts with `amp_`

Author: lutter

docs/amp-powered-subgraphs.md

@@ -41,6 +41,7 @@ The minimum spec version for Amp-powered subgraphs is `1.5.0`.
           - name: Transfer
             file: <IPFS CID of the SQL query file>
 ```
+
 </details>
 
 ### Data source structure
@@ -70,6 +71,7 @@ This is used to assign the subgraph to the appropriate indexing process.
           - name: Transfer
             file: <IPFS CID of the SQL query file>
 ```
+
 </details>
 
 ### `name`
@@ -97,6 +99,7 @@ This name is used for observability purposes and to identify progress and potent
           - name: Transfer
             file: <IPFS CID of the SQL query file>
 ```
+
 </details>
 
 ### `network`
@@ -127,6 +130,7 @@ This is used to validate that the SQL queries for this data source produce resul
           - name: Transfer
             file: <IPFS CID of the SQL query file>
 ```
+
 </details>
 
 ### `source`
@@ -158,6 +162,7 @@ This is used to validate that the SQL queries for this data source only query th
           - name: Transfer
             file: <IPFS CID of the SQL query file>
 ```
+
 </details>
 
 ### `source.tables`
@@ -185,6 +190,7 @@ This is used to validate that the SQL queries for this data source only query th
           - name: Transfer
             file: <IPFS CID of the SQL query file>
 ```
+
 </details>
 
 ### Optional `source.address`
@@ -193,7 +199,7 @@ Contains the contract address with which SQL queries in the data source interact
 
 Enables SQL query reuse through `sg_source_address()` calls instead of hard-coding the contract address.
 SQL queries resolve `sg_source_address()` calls to this contract address.
-    
+
 <details>
 <summary>Example YAML:</summary>
 
@@ -215,6 +221,7 @@ SQL queries resolve `sg_source_address()` calls to this contract address.
           - name: Transfer
             file: <IPFS CID of the SQL query file>
 ```
+
 </details>
 
 ### Optional `source.startBlock`
@@ -245,6 +252,7 @@ _When not provided, defaults to block number `0`._
           - name: Transfer
             file: <IPFS CID of the SQL query file>
 ```
+
 </details>
 
 ### Optional `source.endBlock`
@@ -275,6 +283,7 @@ _When not provided, defaults to the maximum possible block number._
           - name: Transfer
             file: <IPFS CID of the SQL query file>
 ```
+
 </details>
 
 ### `transformer`
@@ -309,6 +318,7 @@ Represents the version of this transformer. Each version may contain a different
           - name: Transfers
             file: <IPFS CID of the SQL query file>
 ```
+
 </details>
 
 ### Optional `transformer.abis`
@@ -344,6 +354,7 @@ _When not provided, defaults to an empty list._
           - name: Transfer
             file: <IPFS CID of the SQL query file>
 ```
+
 </details>
 
 ### `transformer.tables`
@@ -368,6 +379,7 @@ type Block @entity(immutable: true) {
 ```
 
 **YAML manifest:**
+
 ```diff
   specVersion: 1.5.0
 + dataSources:
@@ -386,6 +398,7 @@ type Block @entity(immutable: true) {
 +         - name: Block
             file: <IPFS CID of the SQL query file>
 ```
+
 </details>
 
 ### `transformer.tables[i].query`
@@ -419,6 +432,7 @@ _When not provided, the `file` field is used instead._
           - name: Block
 +           query: SELECT * FROM "edgeandnode/ethereum_mainnet".blocks;
 ```
+
 </details>
 
 ### `transformer.tables[i].file`
@@ -451,6 +465,7 @@ _When not provided, the `query` field is used instead._
           - name: Block
 +           file: <IPFS CID of the SQL query file>
 ```
+
 </details>
 
 ### Amp-powered subgraph examples
@@ -460,6 +475,11 @@ https://github.com/edgeandnode/amp-subgraph-examples
 
 ## SQL query requirements
 
+### Names
+
+The names of tables, columns, and aliases must not start with `amp_` as this
+prefix is reserved for internal use.
+
 ### Block numbers
 
 Every SQL query in Amp-powered subgraphs must return the block number for every row.

graph/src/amp/sql/query_builder/block_range_query.rs

@@ -1,10 +1,8 @@
 use std::{
     collections::BTreeMap,
-    hash::{BuildHasher, Hash, Hasher},
     ops::{ControlFlow, RangeInclusive},
 };
 
-use ahash::RandomState;
 use alloy::primitives::BlockNumber;
 use sqlparser_latest::ast::{self, VisitMut, VisitorMut};
 
@@ -23,11 +21,8 @@ pub(super) fn new_block_range_query(
     block_range: &RangeInclusive<BlockNumber>,
 ) -> ast::Query {
     // CTE names are unique within a SQL query.
-    // The hasher ensures that CTEs created for block range do not collide with user-defined CTEs.
-    // Constant seeds ensure consistent block range queries for the same input parameters.
-    let mut hasher = RandomState::with_seeds(0, 0, 0, 0).build_hasher();
 
-    let tables_to_ctes_mapping = new_tables_to_ctes_mapping(query, &mut hasher);
+    let tables_to_ctes_mapping = new_tables_to_ctes_mapping(query);
     assert!(!tables_to_ctes_mapping.is_empty());
 
     let mut cte_tables = Vec::with_capacity(tables_to_ctes_mapping.len());
@@ -46,24 +41,18 @@ pub(super) fn new_block_range_query(
     let block_range_query = format!(
         "WITH {cte_tables}, {source} AS ({query}) SELECT {source}.* FROM {source} ORDER BY {source}.{block_number_column}",
         cte_tables = cte_tables.join(", "),
-        source = format!("source_{}", hasher.finish())
+        source = format!("amp_src")
     );
 
     parse_query(block_range_query).unwrap()
 }
 
 /// Creates unique CTE names for every table referenced by the SQL query.
-fn new_tables_to_ctes_mapping(
-    query: &ast::Query,
-    hasher: &mut impl Hasher,
-) -> BTreeMap<TableReference, String> {
+fn new_tables_to_ctes_mapping(query: &ast::Query) -> BTreeMap<TableReference, String> {
     extract_tables(query)
         .into_iter()
-        .map(|table| {
-            table.hash(hasher);
-
-            (table, format!("block_range_{}", hasher.finish()))
-        })
+        .enumerate()
+        .map(|(idx, table)| (table, format!("amp_br{}", idx + 1)))
         .collect()
 }
 
@@ -137,18 +126,18 @@ mod tests {
             block_range_query.to_string(),
             parse_query(
                 r#"
-                WITH block_range_14621009630487609643 AS (
+                WITH amp_br1 AS (
                     SELECT * FROM d WHERE _block_num BETWEEN 0 AND 1000000
                 ),
-                source_14621009630487609643 AS (
-                    SELECT a, b, c FROM block_range_14621009630487609643 AS d
+                amp_src AS (
+                    SELECT a, b, c FROM amp_br1 AS d
                 )
                 SELECT
-                    source_14621009630487609643.*
+                    amp_src.*
                 FROM
-                    source_14621009630487609643
+                    amp_src
                 ORDER BY
-                    source_14621009630487609643.b
+                    amp_src.b
                 "#
             )
             .unwrap()
@@ -167,21 +156,21 @@ mod tests {
             block_range_query.to_string(),
             parse_query(
                 r#"
-                WITH block_range_14621009630487609643 AS (
+                WITH amp_br1 AS (
                     SELECT * FROM d WHERE _block_num BETWEEN 0 AND 1000000
                 ),
-                block_range_12377422807768256314 AS (
+                amp_br2 AS (
                     SELECT * FROM e WHERE _block_num BETWEEN 0 AND 1000000
                 ),
-                source_12377422807768256314 AS (
-                    SELECT a, b, c FROM block_range_14621009630487609643 AS d JOIN block_range_12377422807768256314 AS e ON e.e = d.d
+                amp_src AS (
+                    SELECT a, b, c FROM amp_br1 AS d JOIN amp_br2 AS e ON e.e = d.d
                 )
                 SELECT
-                    source_12377422807768256314.*
+                    amp_src.*
                 FROM
-                    source_12377422807768256314
+                    amp_src
                 ORDER BY
-                    source_12377422807768256314.b
+                    amp_src.b
                 "#
             )
             .unwrap()

graph/src/amp/sql/query_builder/context_query.rs

@@ -1,4 +1,3 @@
-use ahash::RandomState;
 use itertools::Itertools;
 use sqlparser_latest::ast;
 
@@ -19,16 +18,11 @@ pub(super) fn new_context_query<'a>(
     context_columns: impl IntoIterator<Item = &'a str>,
 ) -> ast::Query {
     // CTE names are unique within a SQL query.
-    // The hasher ensures that CTEs created for context do not collide with user-defined CTEs.
-    // Constant seeds ensure consistent context queries for the same input parameters.
-    let hasher = RandomState::with_seeds(0, 0, 0, 0);
-    let query_hash = hasher.hash_one(query);
-
     let context_columns = context_columns.into_iter().collect_vec();
     assert!(!context_columns.is_empty());
 
-    let context_cte = format!("context_{query_hash}");
-    let source_cte = format!("source_{query_hash}");
+    let context_cte = "amp_ctx";
+    let source_cte = "amp_src";
 
     let context_query = format!(
         "