Merge pull request #211 from flashbots/faster-writes

Use COPY to speed up database writes for blocks and traces

mev_inspect/crud/blocks.py

@@ -1,6 +1,7 @@
 from datetime import datetime
 from typing import List
 
+from mev_inspect.db import write_as_csv
 from mev_inspect.schemas.blocks import Block
 
 
@@ -28,16 +29,11 @@ def write_blocks(
     db_session,
     blocks: List[Block],
 ) -> None:
-    block_params = [
-        {
-            "block_number": block.block_number,
-            "block_timestamp": datetime.fromtimestamp(block.block_timestamp),
-        }
+    items_generator = (
+        (
+            block.block_number,
+            datetime.fromtimestamp(block.block_timestamp),
+        )
         for block in blocks
-    ]
-
-    db_session.execute(
-        "INSERT INTO blocks (block_number, block_timestamp) VALUES (:block_number, :block_timestamp)",
-        params=block_params,
     )
-    db_session.commit()
+    write_as_csv(db_session, "blocks", items_generator)

mev_inspect/crud/traces.py

@@ -1,6 +1,8 @@
 import json
+from datetime import datetime, timezone
 from typing import List
 
+from mev_inspect.db import to_postgres_list, write_as_csv
 from mev_inspect.models.traces import ClassifiedTraceModel
 from mev_inspect.schemas.traces import ClassifiedTrace
 
@@ -26,30 +28,35 @@ def write_classified_traces(
     db_session,
     classified_traces: List[ClassifiedTrace],
 ) -> None:
-    models = []
-    for trace in classified_traces:
-        inputs_json = (json.loads(trace.json(include={"inputs"}))["inputs"],)
-        models.append(
-            ClassifiedTraceModel(
-                transaction_hash=trace.transaction_hash,
-                transaction_position=trace.transaction_position,
-                block_number=trace.block_number,
-                classification=trace.classification.value,
-                trace_type=trace.type.value,
-                trace_address=trace.trace_address,
-                protocol=str(trace.protocol),
-                abi_name=trace.abi_name,
-                function_name=trace.function_name,
-                function_signature=trace.function_signature,
-                inputs=inputs_json,
-                from_address=trace.from_address,
-                to_address=trace.to_address,
-                gas=trace.gas,
-                value=trace.value,
-                gas_used=trace.gas_used,
-                error=trace.error,
-            )
+    classified_at = datetime.now(timezone.utc)
+    items = (
+        (
+            classified_at,
+            trace.transaction_hash,
+            trace.block_number,
+            trace.classification.value,
+            trace.type.value,
+            str(trace.protocol),
+            trace.abi_name,
+            trace.function_name,
+            trace.function_signature,
+            _inputs_as_json(trace),
+            trace.from_address,
+            trace.to_address,
+            trace.gas,
+            trace.value,
+            trace.gas_used,
+            trace.error,
+            to_postgres_list(trace.trace_address),
+            trace.transaction_position,
         )
+        for trace in classified_traces
+    )
 
-    db_session.bulk_save_objects(models)
-    db_session.commit()
+    write_as_csv(db_session, "classified_traces", items)
+
+
+def _inputs_as_json(trace) -> str:
+    inputs = json.dumps(json.loads(trace.json(include={"inputs"}))["inputs"])
+    inputs_with_array = f"[{inputs}]"
+    return inputs_with_array

mev_inspect/db.py

@@ -1,9 +1,11 @@
 import os
-from typing import Optional
+from typing import Any, Iterable, List, Optional
 
 from sqlalchemy import create_engine, orm
 from sqlalchemy.orm import sessionmaker
 
+from mev_inspect.string_io import StringIteratorIO
+
 
 def get_trace_database_uri() -> Optional[str]:
     username = os.getenv("TRACE_DB_USER")
@@ -63,3 +65,29 @@ def get_trace_session() -> Optional[orm.Session]:
         return Session()
 
     return None
+
+
+def write_as_csv(
+    db_session,
+    table_name: str,
+    items: Iterable[Iterable[Any]],
+) -> None:
+    csv_iterator = StringIteratorIO(
+        ("|".join(map(_clean_csv_value, item)) + "\n" for item in items)
+    )
+
+    with db_session.connection().connection.cursor() as cursor:
+        cursor.copy_from(csv_iterator, table_name, sep="|")
+
+
+def _clean_csv_value(value: Optional[Any]) -> str:
+    if value is None:
+        return r"\N"
+    return str(value).replace("\n", "\\n")
+
+
+def to_postgres_list(values: List[Any]) -> str:
+    if len(values) == 0:
+        return "{}"
+
+    return "{" + ",".join(map(str, values)) + "}"

mev_inspect/string_io.py

@@ -0,0 +1,40 @@
+"""This is taken from https://hakibenita.com/fast-load-data-python-postgresql"""
+
+import io
+from typing import Iterator, Optional
+
+
+class StringIteratorIO(io.TextIOBase):
+    def __init__(self, iter: Iterator[str]):
+        self._iter = iter
+        self._buff = ""
+
+    def readable(self) -> bool:
+        return True
+
+    def _read1(self, n: Optional[int] = None) -> str:
+        while not self._buff:
+            try:
+                self._buff = next(self._iter)
+            except StopIteration:
+                break
+        ret = self._buff[:n]
+        self._buff = self._buff[len(ret) :]
+        return ret
+
+    def read(self, n: Optional[int] = None) -> str:
+        line = []
+        if n is None or n < 0:
+            while True:
+                m = self._read1()
+                if not m:
+                    break
+                line.append(m)
+        else:
+            while n > 0:
+                m = self._read1(n)
+                if not m:
+                    break
+                n -= len(m)
+                line.append(m)
+        return "".join(line)