Convert the runtype classes to attrs

`attrs` is much more beneficial:

* `attrs` is supported by type checkers, such as MyPy & IDEs
* `attrs` is widely used and industry-proven
* `attrs` is explicit in its declarations, there is no magic
* `attrs` has slots

But mainly for the first item — type checking by type checkers.

Those that were runtype classes, are frozen. Those that were not, are unfrozen for now, but we can freeze them later if and where it works (the stricter, the better).

Author: Sergey Vasilyev

data_diff/abcs/compiler.py

@@ -1,9 +1,13 @@
 from abc import ABC
 
+import attrs
 
+
+@attrs.define(frozen=False)
 class AbstractCompiler(ABC):
     pass
 
 
+@attrs.define(frozen=False, eq=False)
 class Compilable(ABC):
     pass

data_diff/abcs/database_types.py

@@ -3,7 +3,7 @@
 from typing import Tuple, Union
 from datetime import datetime
 
-from runtype import dataclass
+import attrs
 
 from data_diff.utils import ArithAlphanumeric, ArithUUID, Unknown
 
@@ -13,14 +13,14 @@
 DbTime = datetime
 
 
-@dataclass
+@attrs.define(frozen=True)
 class ColType:
     @property
     def supported(self) -> bool:
         return True
 
 
-@dataclass
+@attrs.define(frozen=True)
 class PrecisionType(ColType):
     precision: int
     rounds: Union[bool, Unknown] = Unknown
@@ -50,7 +50,7 @@ class Date(TemporalType):
     pass
 
 
-@dataclass
+@attrs.define(frozen=True)
 class NumericType(ColType):
     # 'precision' signifies how many fractional digits (after the dot) we want to compare
     precision: int
@@ -84,7 +84,7 @@ def python_type(self) -> type:
         return decimal.Decimal
 
 
-@dataclass
+@attrs.define(frozen=True)
 class StringType(ColType):
     python_type = str
 
@@ -122,7 +122,7 @@ class String_VaryingAlphanum(String_Alphanum):
     pass
 
 
-@dataclass
+@attrs.define(frozen=True)
 class String_FixedAlphanum(String_Alphanum):
     length: int
 
@@ -132,20 +132,20 @@ def make_value(self, value):
         return self.python_type(value, max_len=self.length)
 
 
-@dataclass
+@attrs.define(frozen=True)
 class Text(StringType):
     @property
     def supported(self) -> bool:
         return False
 
 
 # In majority of DBMSes, it is called JSON/JSONB. Only in Snowflake, it is OBJECT.
-@dataclass
+@attrs.define(frozen=True)
 class JSON(ColType):
     pass
 
 
-@dataclass
+@attrs.define(frozen=True)
 class Array(ColType):
     item_type: ColType
 
@@ -155,21 +155,21 @@ class Array(ColType):
 # For example, in BigQuery:
 # - https://cloud.google.com/bigquery/docs/reference/standard-sql/data-types#struct_type
 # - https://cloud.google.com/bigquery/docs/reference/standard-sql/lexical#struct_literals
-@dataclass
+@attrs.define(frozen=True)
 class Struct(ColType):
     pass
 
 
-@dataclass
+@attrs.define(frozen=True)
 class Integer(NumericType, IKey):
     precision: int = 0
     python_type: type = int
 
-    def __post_init__(self):
+    def __attrs_post_init__(self):
         assert self.precision == 0
 
 
-@dataclass
+@attrs.define(frozen=True)
 class UnknownColType(ColType):
     text: str
 

data_diff/cloud/datafold_api.py

@@ -1,9 +1,9 @@
 import base64
-import dataclasses
 import enum
 import time
 from typing import Any, Dict, List, Optional, Type, Tuple
 
+import attrs
 import pydantic
 import requests
 from typing_extensions import Self
@@ -178,13 +178,13 @@ class TCloudApiDataSourceTestResult(pydantic.BaseModel):
     result: Optional[TCloudDataSourceTestResult]
 
 
-@dataclasses.dataclass
+@attrs.define(frozen=True)
 class DatafoldAPI:
     api_key: str
     host: str = "https://app.datafold.com"
     timeout: int = 30
 
-    def __post_init__(self):
+    def __attrs_post_init__(self):
         self.host = self.host.rstrip("/")
         self.headers = {
             "Authorization": f"Key {self.api_key}",

data_diff/databases/_connect.py

@@ -3,10 +3,11 @@
 from itertools import zip_longest
 from contextlib import suppress
 import weakref
+
+import attrs
 import dsnparse
 import toml
 
-from runtype import dataclass
 from typing_extensions import Self
 
 from data_diff.databases.base import Database, ThreadedDatabase
@@ -25,7 +26,7 @@
 from data_diff.databases.mssql import MsSQL
 
 
-@dataclass
+@attrs.define(frozen=True)
 class MatchUriPath:
     database_cls: Type[Database]
 

data_diff/databases/base.py

@@ -1,6 +1,6 @@
 import abc
 import functools
-from dataclasses import field
+import random
 from datetime import datetime
 import math
 import sys
@@ -14,7 +14,7 @@
 import decimal
 import contextvars
 
-from runtype import dataclass
+import attrs
 from typing_extensions import Self
 
 from data_diff.abcs.compiler import AbstractCompiler
@@ -90,12 +90,7 @@ class CompileError(Exception):
     pass
 
 
-# TODO: remove once switched to attrs, where ForwardRef[]/strings are resolved.
-class _RuntypeHackToFixCicularRefrencedDatabase:
-    dialect: "BaseDialect"
-
-
-@dataclass
+@attrs.define(frozen=True)
 class Compiler(AbstractCompiler):
     """
     Compiler bears the context for a single compilation.
@@ -107,16 +102,16 @@ class Compiler(AbstractCompiler):
     # Database is needed to normalize tables. Dialect is needed for recursive compilations.
     # In theory, it is many-to-many relations: e.g. a generic ODBC driver with multiple dialects.
     # In practice, we currently bind the dialects to the specific database classes.
-    database: _RuntypeHackToFixCicularRefrencedDatabase
+    database: "Database"
 
     in_select: bool = False  # Compilation runtime flag
     in_join: bool = False  # Compilation runtime flag
 
-    _table_context: List = field(default_factory=list)  # List[ITable]
-    _subqueries: Dict[str, Any] = field(default_factory=dict)  # XXX not thread-safe
+    _table_context: List = attrs.field(factory=list)  # List[ITable]
+    _subqueries: Dict[str, Any] = attrs.field(factory=dict)  # XXX not thread-safe
     root: bool = True
 
-    _counter: List = field(default_factory=lambda: [0])
+    _counter: List = attrs.field(factory=lambda: [0])
 
     @property
     def dialect(self) -> "BaseDialect":
@@ -136,7 +131,7 @@ def new_unique_table_name(self, prefix="tmp") -> DbPath:
         return self.database.dialect.parse_table_name(table_name)
 
     def add_table_context(self, *tables: Sequence, **kw) -> Self:
-        return self.replace(_table_context=self._table_context + list(tables), **kw)
+        return attrs.evolve(self, table_context=self._table_context + list(tables), **kw)
 
 
 def parse_table_name(t):
@@ -271,7 +266,7 @@ def _compile(self, compiler: Compiler, elem) -> str:
         if elem is None:
             return "NULL"
         elif isinstance(elem, Compilable):
-            return self.render_compilable(compiler.replace(root=False), elem)
+            return self.render_compilable(attrs.evolve(compiler, root=False), elem)
         elif isinstance(elem, str):
             return f"'{elem}'"
         elif isinstance(elem, (int, float)):
@@ -381,7 +376,7 @@ def render_column(self, c: Compiler, elem: Column) -> str:
         return self.quote(elem.name)
 
     def render_cte(self, parent_c: Compiler, elem: Cte) -> str:
-        c: Compiler = parent_c.replace(_table_context=[], in_select=False)
+        c: Compiler = attrs.evolve(parent_c, table_context=[], in_select=False)
         compiled = self.compile(c, elem.source_table)
 
         name = elem.name or parent_c.new_unique_name()
@@ -494,7 +489,7 @@ def render_tablealias(self, c: Compiler, elem: TableAlias) -> str:
         return f"{self.compile(c, elem.source_table)} {self.quote(elem.name)}"
 
     def render_tableop(self, parent_c: Compiler, elem: TableOp) -> str:
-        c: Compiler = parent_c.replace(in_select=False)
+        c: Compiler = attrs.evolve(parent_c, in_select=False)
         table_expr = f"{self.compile(c, elem.table1)} {elem.op} {self.compile(c, elem.table2)}"
         if parent_c.in_select:
             table_expr = f"({table_expr}) {c.new_unique_name()}"
@@ -506,7 +501,7 @@ def render__resolvecolumn(self, c: Compiler, elem: _ResolveColumn) -> str:
         return self.compile(c, elem._get_resolved())
 
     def render_select(self, parent_c: Compiler, elem: Select) -> str:
-        c: Compiler = parent_c.replace(in_select=True)  # .add_table_context(self.table)
+        c: Compiler = attrs.evolve(parent_c, in_select=True)  # .add_table_context(self.table)
         compile_fn = functools.partial(self.compile, c)
 
         columns = ", ".join(map(compile_fn, elem.columns)) if elem.columns else "*"
@@ -544,7 +539,8 @@ def render_select(self, parent_c: Compiler, elem: Select) -> str:
 
     def render_join(self, parent_c: Compiler, elem: Join) -> str:
         tables = [
-            t if isinstance(t, TableAlias) else TableAlias(t, parent_c.new_unique_name()) for t in elem.source_tables
+            t if isinstance(t, TableAlias) else TableAlias(source_table=t, name=parent_c.new_unique_name())
+            for t in elem.source_tables
         ]
         c = parent_c.add_table_context(*tables, in_join=True, in_select=False)
         op = " JOIN " if elem.op is None else f" {elem.op} JOIN "
@@ -577,7 +573,8 @@ def render_groupby(self, c: Compiler, elem: GroupBy) -> str:
         if isinstance(elem.table, Select) and elem.table.columns is None and elem.table.group_by_exprs is None:
             return self.compile(
                 c,
-                elem.table.replace(
+                attrs.evolve(
+                    elem.table,
                     columns=columns,
                     group_by_exprs=[Code(k) for k in keys],
                     having_exprs=elem.having_exprs,
@@ -589,7 +586,7 @@ def render_groupby(self, c: Compiler, elem: GroupBy) -> str:
         having_str = (
             " HAVING " + " AND ".join(map(compile_fn, elem.having_exprs)) if elem.having_exprs is not None else ""
         )
-        select = f"SELECT {columns_str} FROM {self.compile(c.replace(in_select=True), elem.table)} GROUP BY {keys_str}{having_str}"
+        select = f"SELECT {columns_str} FROM {self.compile(attrs.evolve(c, in_select=True), elem.table)} GROUP BY {keys_str}{having_str}"
 
         if c.in_select:
             select = f"({select}) {c.new_unique_name()}"
@@ -815,7 +812,7 @@ def set_timezone_to_utc(self) -> str:
 T = TypeVar("T", bound=BaseDialect)
 
 
-@dataclass
+@attrs.define(frozen=True)
 class QueryResult:
     rows: list
     columns: Optional[list] = None
@@ -830,7 +827,7 @@ def __getitem__(self, i):
         return self.rows[i]
 
 
-class Database(abc.ABC, _RuntypeHackToFixCicularRefrencedDatabase):
+class Database(abc.ABC):
     """Base abstract class for databases.
 
     Used for providing connection code and implementation specific SQL utilities.

data_diff/diff_tables.py

@@ -3,14 +3,13 @@
 
 import time
 from abc import ABC, abstractmethod
-from dataclasses import field
 from enum import Enum
 from contextlib import contextmanager
 from operator import methodcaller
 from typing import Dict, Tuple, Iterator, Optional
 from concurrent.futures import ThreadPoolExecutor, as_completed
 
-from runtype import dataclass
+import attrs
 
 from data_diff.info_tree import InfoTree, SegmentInfo
 from data_diff.utils import dbt_diff_string_template, run_as_daemon, safezip, getLogger, truncate_error, Vector
@@ -31,7 +30,7 @@ class Algorithm(Enum):
 DiffResult = Iterator[Tuple[str, tuple]]  # Iterator[Tuple[Literal["+", "-"], tuple]]
 
 
-@dataclass
+@attrs.define(frozen=True)
 class ThreadBase:
     "Provides utility methods for optional threading"
 
@@ -72,7 +71,7 @@ def _run_in_background(self, *funcs):
                 f.result()
 
 
-@dataclass
+@attrs.define(frozen=True)
 class DiffStats:
     diff_by_sign: Dict[str, int]
     table1_count: int
@@ -82,12 +81,12 @@ class DiffStats:
     extra_column_diffs: Optional[Dict[str, int]]
 
 
-@dataclass
+@attrs.define(frozen=True)
 class DiffResultWrapper:
     diff: iter  # DiffResult
     info_tree: InfoTree
     stats: dict
-    result_list: list = field(default_factory=list)
+    result_list: list = attrs.field(factory=list)
 
     def __iter__(self):
         yield from self.result_list
@@ -203,7 +202,7 @@ def diff_tables(self, table1: TableSegment, table2: TableSegment, info_tree: Inf
 
     def _diff_tables_wrapper(self, table1: TableSegment, table2: TableSegment, info_tree: InfoTree) -> DiffResult:
         if is_tracking_enabled():
-            options = dict(self)
+            options = attrs.asdict(self, recurse=False)
             options["differ_name"] = type(self).__name__
             event_json = create_start_event_json(options)
             run_as_daemon(send_event_json, event_json)