datachain 0.3.17__py3-none-any.whl → 0.3.19__py3-none-any.whl

datachain/__init__.py

@@ -1,21 +1,23 @@
 from datachain.lib.data_model import DataModel, DataType, is_chain_type
 from datachain.lib.dc import C, Column, DataChain, Sys
 from datachain.lib.file import (
+    ArrowRow,
     File,
     FileError,
     ImageFile,
-    IndexedFile,
     TarVFile,
     TextFile,
 )
 from datachain.lib.model_store import ModelStore
 from datachain.lib.udf import Aggregator, Generator, Mapper
 from datachain.lib.utils import AbstractUDF, DataChainError
+from datachain.query import metrics, param
 from datachain.query.session import Session
 
 __all__ = [
     "AbstractUDF",
     "Aggregator",
+    "ArrowRow",
     "C",
     "Column",
     "DataChain",
@@ -26,7 +28,6 @@ __all__ = [
     "FileError",
     "Generator",
     "ImageFile",
-    "IndexedFile",
     "Mapper",
     "ModelStore",
     "Session",
@@ -34,4 +35,6 @@ __all__ = [
     "TarVFile",
     "TextFile",
     "is_chain_type",
+    "metrics",
+    "param",
 ]

datachain/cache.py

@@ -1,56 +1,15 @@
-import hashlib
-import json
 import os
-from datetime import datetime
-from functools import partial
 from typing import TYPE_CHECKING, Optional
 
-import attrs
 from dvc_data.hashfile.db.local import LocalHashFileDB
 from dvc_objects.fs.local import LocalFileSystem
 from fsspec.callbacks import Callback, TqdmCallback
 
-from datachain.utils import TIME_ZERO
-
 from .progress import Tqdm
 
 if TYPE_CHECKING:
     from datachain.client import Client
-    from datachain.storage import StorageURI
-
-sha256 = partial(hashlib.sha256, usedforsecurity=False)
-
-
-@attrs.frozen
-class UniqueId:
-    storage: "StorageURI"
-    path: str
-    size: int
-    etag: str
-    version: str = ""
-    is_latest: bool = True
-    location: Optional[str] = None
-    last_modified: datetime = TIME_ZERO
-
-    def get_parsed_location(self) -> Optional[dict]:
-        if not self.location:
-            return None
-
-        loc_stack = (
-            json.loads(self.location)
-            if isinstance(self.location, str)
-            else self.location
-        )
-        if len(loc_stack) > 1:
-            raise NotImplementedError("Nested v-objects are not supported yet.")
-
-        return loc_stack[0]
-
-    def get_hash(self) -> str:
-        fingerprint = f"{self.storage}/{self.path}/{self.version}/{self.etag}"
-        if self.location:
-            fingerprint += f"/{self.location}"
-        return sha256(fingerprint.encode()).hexdigest()
+    from datachain.lib.file import File
 
 
 def try_scandir(path):
@@ -77,30 +36,30 @@ class DataChainCache:
     def tmp_dir(self):
         return self.odb.tmp_dir
 
-    def get_path(self, uid: UniqueId) -> Optional[str]:
-        if self.contains(uid):
-            return self.path_from_checksum(uid.get_hash())
+    def get_path(self, file: "File") -> Optional[str]:
+        if self.contains(file):
+            return self.path_from_checksum(file.get_hash())
         return None
 
-    def contains(self, uid: UniqueId) -> bool:
-        return self.odb.exists(uid.get_hash())
+    def contains(self, file: "File") -> bool:
+        return self.odb.exists(file.get_hash())
 
     def path_from_checksum(self, checksum: str) -> str:
         assert checksum
         return self.odb.oid_to_path(checksum)
 
-    def remove(self, uid: UniqueId) -> None:
-        self.odb.delete(uid.get_hash())
+    def remove(self, file: "File") -> None:
+        self.odb.delete(file.get_hash())
 
     async def download(
-        self, uid: UniqueId, client: "Client", callback: Optional[Callback] = None
+        self, file: "File", client: "Client", callback: Optional[Callback] = None
     ) -> None:
-        from_path = f"{uid.storage}/{uid.path}"
+        from_path = f"{file.source}/{file.path}"
         from dvc_objects.fs.utils import tmp_fname
 
         odb_fs = self.odb.fs
         tmp_info = odb_fs.join(self.odb.tmp_dir, tmp_fname())  # type: ignore[arg-type]
-        size = uid.size
+        size = file.size
         if size < 0:
             size = await client.get_size(from_path)
         cb = callback or TqdmCallback(
@@ -115,13 +74,13 @@ class DataChainCache:
                 cb.close()
 
         try:
-            oid = uid.get_hash()
+            oid = file.get_hash()
             self.odb.add(tmp_info, self.odb.fs, oid)
         finally:
             os.unlink(tmp_info)
 
-    def store_data(self, uid: UniqueId, contents: bytes) -> None:
-        checksum = uid.get_hash()
+    def store_data(self, file: "File", contents: bytes) -> None:
+        checksum = file.get_hash()
         dst = self.path_from_checksum(checksum)
         if not os.path.exists(dst):
             # Create the file only if it's not already in cache

datachain/catalog/catalog.py

@@ -1,4 +1,3 @@
-import ast
 import glob
 import io
 import json
@@ -34,7 +33,7 @@ import yaml
 from sqlalchemy import Column
 from tqdm import tqdm
 
-from datachain.cache import DataChainCache, UniqueId
+from datachain.cache import DataChainCache
 from datachain.client import Client
 from datachain.config import get_remote_config, read_config
 from datachain.dataset import (
@@ -53,9 +52,9 @@ from datachain.error import (
     DataChainError,
     DatasetInvalidVersionError,
     DatasetNotFoundError,
+    DatasetVersionNotFoundError,
     PendingIndexingError,
     QueryScriptCancelError,
-    QueryScriptCompileError,
     QueryScriptRunError,
 )
 from datachain.listing import Listing
@@ -588,44 +587,13 @@ class Catalog:
     def generate_query_dataset_name(cls) -> str:
         return f"{QUERY_DATASET_PREFIX}_{uuid4().hex}"
 
-    def attach_query_wrapper(self, code_ast):
-        if code_ast.body:
-            last_expr = code_ast.body[-1]
-            if isinstance(last_expr, ast.Expr):
-                new_expressions = [
-                    ast.Import(
-                        names=[ast.alias(name="datachain.query.dataset", asname=None)]
-                    ),
-                    ast.Expr(
-                        value=ast.Call(
-                            func=ast.Attribute(
-                                value=ast.Attribute(
-                                    value=ast.Attribute(
-                                        value=ast.Name(id="datachain", ctx=ast.Load()),
-                                        attr="query",
-                                        ctx=ast.Load(),
-                                    ),
-                                    attr="dataset",
-                                    ctx=ast.Load(),
-                                ),
-                                attr="query_wrapper",
-                                ctx=ast.Load(),
-                            ),
-                            args=[last_expr],
-                            keywords=[],
-                        )
-                    ),
-                ]
-                code_ast.body[-1:] = new_expressions
-        return code_ast
-
-    def get_client(self, uri: StorageURI, **config: Any) -> Client:
+    def get_client(self, uri: str, **config: Any) -> Client:
         """
         Return the client corresponding to the given source `uri`.
         """
         config = config or self.client_config
         cls = Client.get_implementation(uri)
-        return cls.from_source(uri, self.cache, **config)
+        return cls.from_source(StorageURI(uri), self.cache, **config)
 
     def enlist_source(
         self,
@@ -1218,7 +1186,9 @@ class Catalog:
 
         dataset_version = dataset.get_version(version)
         if not dataset_version:
-            raise ValueError(f"Dataset {dataset.name} does not have version {version}")
+            raise DatasetVersionNotFoundError(
+                f"Dataset {dataset.name} does not have version {version}"
+            )
 
         if not dataset_version.is_final_status():
             raise ValueError("Cannot register dataset version in non final status")
@@ -1431,7 +1401,7 @@ class Catalog:
 
     def get_file_signals(
         self, dataset_name: str, dataset_version: int, row: RowDict
-    ) -> Optional[dict]:
+    ) -> Optional[RowDict]:
         """
         Function that returns file signals from dataset row.
         Note that signal names are without prefix, so if there was 'laion__file__source'
@@ -1448,7 +1418,7 @@ class Catalog:
 
         version = self.get_dataset(dataset_name).get_version(dataset_version)
 
-        file_signals_values = {}
+        file_signals_values = RowDict()
 
         schema = SignalSchema.deserialize(version.feature_schema)
         for file_signals in schema.get_signals(File):
@@ -1476,6 +1446,8 @@ class Catalog:
         use_cache: bool = True,
         **config: Any,
     ):
+        from datachain.lib.file import File
+
         file_signals = self.get_file_signals(dataset_name, dataset_version, row)
         if not file_signals:
             raise RuntimeError("Cannot open object without file signals")
@@ -1483,22 +1455,10 @@ class Catalog:
         config = config or self.client_config
         client = self.get_client(file_signals["source"], **config)
         return client.open_object(
-            self._get_row_uid(file_signals),  # type: ignore [arg-type]
+            File._from_row(file_signals),
             use_cache=use_cache,
         )
 
-    def _get_row_uid(self, row: RowDict) -> UniqueId:
-        return UniqueId(
-            row["source"],
-            row["path"],
-            row["size"],
-            row["etag"],
-            row["version"],
-            row["is_latest"],
-            row["location"],
-            row["last_modified"],
-        )
-
     def ls(
         self,
         sources: list[str],
@@ -1591,7 +1551,7 @@ class Catalog:
 
         try:
             remote_dataset_version = remote_dataset.get_version(version)
-        except (ValueError, StopIteration) as exc:
+        except (DatasetVersionNotFoundError, StopIteration) as exc:
             raise DataChainError(
                 f"Dataset {remote_dataset_name} doesn't have version {version}"
                 " on server"
@@ -1732,64 +1692,24 @@ class Catalog:
         query_script: str,
         env: Optional[Mapping[str, str]] = None,
         python_executable: str = sys.executable,
-        save: bool = False,
-        capture_output: bool = True,
+        capture_output: bool = False,
         output_hook: Callable[[str], None] = noop,
         params: Optional[dict[str, str]] = None,
         job_id: Optional[str] = None,
-        _execute_last_expression: bool = False,
     ) -> None:
-        """
-        Method to run custom user Python script to run a query and, as result,
-        creates new dataset from the results of a query.
-        Returns tuple of result dataset and script output.
-
-        Constraints on query script:
-            1. datachain.query.DatasetQuery should be used in order to create query
-            for a dataset
-            2. There should not be any .save() call on DatasetQuery since the idea
-            is to create only one dataset as the outcome of the script
-            3. Last statement must be an instance of DatasetQuery
-
-        If save is set to True, we are creating new dataset with results
-        from dataset query. If it's set to False, we will just print results
-        without saving anything
-
-        Example of query script:
-            from datachain.query import DatasetQuery, C
-            DatasetQuery('s3://ldb-public/remote/datasets/mnist-tiny/').filter(
-                C.size > 1000
-            )
-        """
-        if _execute_last_expression:
-            try:
-                code_ast = ast.parse(query_script)
-                code_ast = self.attach_query_wrapper(code_ast)
-                query_script_compiled = ast.unparse(code_ast)
-            except Exception as exc:
-                raise QueryScriptCompileError(
-                    f"Query script failed to compile, reason: {exc}"
-                ) from exc
-        else:
-            query_script_compiled = query_script
-            assert not save
-
+        cmd = [python_executable, "-c", query_script]
         env = dict(env or os.environ)
         env.update(
             {
                 "DATACHAIN_QUERY_PARAMS": json.dumps(params or {}),
-                "PYTHONPATH": os.getcwd(),  # For local imports
-                "DATACHAIN_QUERY_SAVE": "1" if save else "",
-                "PYTHONUNBUFFERED": "1",
                 "DATACHAIN_JOB_ID": job_id or "",
             },
         )
-        popen_kwargs = {}
+        popen_kwargs: dict[str, Any] = {}
         if capture_output:
             popen_kwargs = {"stdout": subprocess.PIPE, "stderr": subprocess.STDOUT}
 
-        cmd = [python_executable, "-c", query_script_compiled]
-        with subprocess.Popen(cmd, env=env, **popen_kwargs) as proc:  # type: ignore[call-overload]  # noqa: S603
+        with subprocess.Popen(cmd, env=env, **popen_kwargs) as proc:  # noqa: S603
             if capture_output:
                 args = (proc.stdout, output_hook)
                 thread = Thread(target=_process_stream, args=args, daemon=True)

datachain/cli.py

@@ -15,6 +15,7 @@ import shtab
 from datachain import utils
 from datachain.cli_utils import BooleanOptionalAction, CommaSeparatedArgs, KeyValueArgs
 from datachain.lib.dc import DataChain
+from datachain.telemetry import telemetry
 from datachain.utils import DataChainDir
 
 if TYPE_CHECKING:
@@ -803,7 +804,6 @@ def query(
         catalog.query(
             script_content,
             python_executable=python_executable,
-            capture_output=False,
             params=params,
             job_id=job_id,
         )
@@ -872,6 +872,7 @@ def main(argv: Optional[list[str]] = None) -> int:  # noqa: C901, PLR0912, PLR09
         # This also sets this environment variable for any subprocesses
         os.environ["DEBUG_SHOW_SQL_QUERIES"] = "True"
 
+    error = None
     try:
         catalog = get_catalog(client_config=client_config)
         if args.command == "cp":
@@ -1003,14 +1004,16 @@ def main(argv: Optional[list[str]] = None) -> int:  # noqa: C901, PLR0912, PLR09
             print(f"invalid command: {args.command}", file=sys.stderr)
             return 1
         return 0
-    except BrokenPipeError:
+    except BrokenPipeError as exc:
         # Python flushes standard streams on exit; redirect remaining output
         # to devnull to avoid another BrokenPipeError at shutdown
         # See: https://docs.python.org/3/library/signal.html#note-on-sigpipe
+        error = str(exc)
         devnull = os.open(os.devnull, os.O_WRONLY)
         os.dup2(devnull, sys.stdout.fileno())
         return 141  # 128 + 13 (SIGPIPE)
     except (KeyboardInterrupt, Exception) as exc:
+        error = str(exc)
         if isinstance(exc, KeyboardInterrupt):
             msg = "Operation cancelled by the user"
         else:
@@ -1028,3 +1031,5 @@ def main(argv: Optional[list[str]] = None) -> int:  # noqa: C901, PLR0912, PLR09
 
             pdb.post_mortem()
         return 1
+    finally:
+        telemetry.send_cli_call(args.command, error=error)

datachain/client/fsspec.py

@@ -3,7 +3,6 @@ import functools
 import logging
 import multiprocessing
 import os
-import posixpath
 import re
 import sys
 from abc import ABC, abstractmethod
@@ -26,8 +25,8 @@ from fsspec.asyn import get_loop, sync
 from fsspec.callbacks import DEFAULT_CALLBACK, Callback
 from tqdm import tqdm
 
-from datachain.cache import DataChainCache, UniqueId
-from datachain.client.fileslice import FileSlice, FileWrapper
+from datachain.cache import DataChainCache
+from datachain.client.fileslice import FileWrapper
 from datachain.error import ClientError as DataChainClientError
 from datachain.lib.file import File
 from datachain.nodes_fetcher import NodesFetcher
@@ -187,8 +186,8 @@ class Client(ABC):
     def url(self, path: str, expires: int = 3600, **kwargs) -> str:
         return self.fs.sign(self.get_full_path(path), expiration=expires, **kwargs)
 
-    async def get_current_etag(self, uid: UniqueId) -> str:
-        info = await self.fs._info(self.get_full_path(uid.path))
+    async def get_current_etag(self, file: "File") -> str:
+        info = await self.fs._info(self.get_full_path(file.path))
         return self.info_to_file(info, "").etag
 
     async def get_size(self, path: str) -> int:
@@ -317,7 +316,7 @@ class Client(ABC):
 
     def instantiate_object(
         self,
-        uid: UniqueId,
+        file: "File",
         dst: str,
         progress_bar: tqdm,
         force: bool = False,
@@ -328,10 +327,10 @@ class Client(ABC):
             else:
                 progress_bar.close()
                 raise FileExistsError(f"Path {dst} already exists")
-        self.do_instantiate_object(uid, dst)
+        self.do_instantiate_object(file, dst)
 
-    def do_instantiate_object(self, uid: "UniqueId", dst: str) -> None:
-        src = self.cache.get_path(uid)
+    def do_instantiate_object(self, file: "File", dst: str) -> None:
+        src = self.cache.get_path(file)
         assert src is not None
 
         try:
@@ -341,66 +340,33 @@ class Client(ABC):
             copy2(src, dst)
 
     def open_object(
-        self, uid: UniqueId, use_cache: bool = True, cb: Callback = DEFAULT_CALLBACK
+        self, file: File, use_cache: bool = True, cb: Callback = DEFAULT_CALLBACK
     ) -> BinaryIO:
         """Open a file, including files in tar archives."""
-        location = uid.get_parsed_location()
-        if use_cache and (cache_path := self.cache.get_path(uid)):
+        if use_cache and (cache_path := self.cache.get_path(file)):
             return open(cache_path, mode="rb")  # noqa: SIM115
-        if location and location["vtype"] == "tar":
-            return self._open_tar(uid, use_cache=True)
-        return FileWrapper(self.fs.open(self.get_full_path(uid.path)), cb)  # type: ignore[return-value]
-
-    def _open_tar(self, uid: UniqueId, use_cache: bool = True):
-        location = uid.get_parsed_location()
-        assert location
-
-        offset = location["offset"]
-        size = location["size"]
-        parent = location["parent"]
-
-        parent_uid = UniqueId(
-            parent["source"],
-            parent["path"],
-            parent["size"],
-            parent["etag"],
-            location=parent["location"],
-        )
-        f = self.open_object(parent_uid, use_cache=use_cache)
-        return FileSlice(f, offset, size, posixpath.basename(uid.path))
-
-    def download(self, uid: UniqueId, *, callback: Callback = DEFAULT_CALLBACK) -> None:
-        sync(get_loop(), functools.partial(self._download, uid, callback=callback))
-
-    async def _download(self, uid: UniqueId, *, callback: "Callback" = None) -> None:
-        if self.cache.contains(uid):
+        assert not file.location
+        return FileWrapper(self.fs.open(self.get_full_path(file.path)), cb)  # type: ignore[return-value]
+
+    def download(self, file: File, *, callback: Callback = DEFAULT_CALLBACK) -> None:
+        sync(get_loop(), functools.partial(self._download, file, callback=callback))
+
+    async def _download(self, file: File, *, callback: "Callback" = None) -> None:
+        if self.cache.contains(file):
             # Already in cache, so there's nothing to do.
             return
-        await self._put_in_cache(uid, callback=callback)
+        await self._put_in_cache(file, callback=callback)
 
-    def put_in_cache(self, uid: UniqueId, *, callback: "Callback" = None) -> None:
-        sync(get_loop(), functools.partial(self._put_in_cache, uid, callback=callback))
+    def put_in_cache(self, file: File, *, callback: "Callback" = None) -> None:
+        sync(get_loop(), functools.partial(self._put_in_cache, file, callback=callback))
 
-    async def _put_in_cache(
-        self, uid: UniqueId, *, callback: "Callback" = None
-    ) -> None:
-        location = uid.get_parsed_location()
-        if location and location["vtype"] == "tar":
-            loop = asyncio.get_running_loop()
-            await loop.run_in_executor(
-                None, functools.partial(self._download_from_tar, uid, callback=callback)
-            )
-            return
-        if uid.etag:
-            etag = await self.get_current_etag(uid)
-            if uid.etag != etag:
+    async def _put_in_cache(self, file: File, *, callback: "Callback" = None) -> None:
+        assert not file.location
+        if file.etag:
+            etag = await self.get_current_etag(file)
+            if file.etag != etag:
                 raise FileNotFoundError(
-                    f"Invalid etag for {uid.storage}/{uid.path}: "
-                    f"expected {uid.etag}, got {etag}"
+                    f"Invalid etag for {file.source}/{file.path}: "
+                    f"expected {file.etag}, got {etag}"
                 )
-        await self.cache.download(uid, self, callback=callback)
-
-    def _download_from_tar(self, uid, *, callback: "Callback" = None):
-        with self._open_tar(uid, use_cache=False) as f:
-            contents = f.read()
-        self.cache.store_data(uid, contents)
+        await self.cache.download(file, self, callback=callback)

datachain/client/local.py

@@ -7,7 +7,6 @@ from urllib.parse import urlparse
 
 from fsspec.implementations.local import LocalFileSystem
 
-from datachain.cache import UniqueId
 from datachain.lib.file import File
 from datachain.storage import StorageURI
 
@@ -114,8 +113,8 @@ class FileClient(Client):
             use_symlinks=use_symlinks,
         )
 
-    async def get_current_etag(self, uid: UniqueId) -> str:
-        info = self.fs.info(self.get_full_path(uid.path))
+    async def get_current_etag(self, file: "File") -> str:
+        info = self.fs.info(self.get_full_path(file.path))
         return self.info_to_file(info, "").etag
 
     async def get_size(self, path: str) -> int:

datachain/dataset.py

@@ -12,6 +12,7 @@ from typing import (
 from urllib.parse import urlparse
 
 from datachain.client import Client
+from datachain.error import DatasetVersionNotFoundError
 from datachain.sql.types import NAME_TYPES_MAPPING, SQLType
 
 if TYPE_CHECKING:
@@ -417,7 +418,9 @@ class DatasetRecord:
 
     def get_version(self, version: int) -> DatasetVersion:
         if not self.has_version(version):
-            raise ValueError(f"Dataset {self.name} does not have version {version}")
+            raise DatasetVersionNotFoundError(
+                f"Dataset {self.name} does not have version {version}"
+            )
         return next(
             v
             for v in self.versions  # type: ignore [union-attr]
@@ -435,7 +438,9 @@ class DatasetRecord:
         Get identifier in the form my-dataset@v3
         """
         if not self.has_version(version):
-            raise ValueError(f"Dataset {self.name} doesn't have a version {version}")
+            raise DatasetVersionNotFoundError(
+                f"Dataset {self.name} doesn't have a version {version}"
+            )
         return f"{self.name}@v{version}"
 
     def uri(self, version: int) -> str:

datachain/error.py

@@ -10,6 +10,10 @@ class DatasetNotFoundError(NotFoundError):
     pass
 
 
+class DatasetVersionNotFoundError(NotFoundError):
+    pass
+
+
 class DatasetInvalidVersionError(Exception):
     pass
 
@@ -32,14 +36,12 @@ class QueryScriptRunError(Exception):
     Attributes:
         message      Explanation of the error
         return_code  Code returned by the subprocess
-        output       STDOUT + STDERR output of the subprocess
     """
 
-    def __init__(self, message: str, return_code: int = 0, output: str = ""):
+    def __init__(self, message: str, return_code: int = 0):
         self.message = message
         self.return_code = return_code
-        self.output = output
-        super().__init__(self.message)
+        super().__init__(message)
 
 
 class QueryScriptCancelError(QueryScriptRunError):

datachain/lib/arrow.py

@@ -4,11 +4,11 @@ from tempfile import NamedTemporaryFile
 from typing import TYPE_CHECKING, Optional
 
 import pyarrow as pa
-from pyarrow.dataset import dataset
+from pyarrow.dataset import CsvFileFormat, dataset
 from tqdm import tqdm
 
 from datachain.lib.data_model import dict_to_data_model
-from datachain.lib.file import File, IndexedFile
+from datachain.lib.file import ArrowRow, File
 from datachain.lib.model_store import ModelStore
 from datachain.lib.udf import Generator
 
@@ -49,7 +49,8 @@ class ArrowGenerator(Generator):
 
     def process(self, file: File):
         if file._caching_enabled:
-            path = file.get_local_path(download=True)
+            file.ensure_cached()
+            path = file.get_local_path()
             ds = dataset(path, schema=self.input_schema, **self.kwargs)
         elif self.nrows:
             path = _nrows_file(file, self.nrows)
@@ -83,7 +84,12 @@ class ArrowGenerator(Generator):
                                 vals_dict[field] = val
                         vals = [self.output_schema(**vals_dict)]
                     if self.source:
-                        yield [IndexedFile(file=file, index=index), *vals]
+                        kwargs: dict = self.kwargs
+                        # Can't serialize CsvFileFormat; may lose formatting options.
+                        if isinstance(kwargs.get("format"), CsvFileFormat):
+                            kwargs["format"] = "csv"
+                        arrow_file = ArrowRow(file=file, index=index, kwargs=kwargs)
+                        yield [arrow_file, *vals]
                     else:
                         yield vals
                     index += 1