pixeltable 0.4.17__py3-none-any.whl → 0.4.18__py3-none-any.whl

pixeltable/functions/video.py

@@ -306,7 +306,14 @@ def _handle_ffmpeg_error(e: subprocess.CalledProcessError) -> NoReturn:
 
 @pxt.udf(is_method=True)
 def clip(
-    video: pxt.Video, *, start_time: float, end_time: float | None = None, duration: float | None = None
+    video: pxt.Video,
+    *,
+    start_time: float,
+    end_time: float | None = None,
+    duration: float | None = None,
+    mode: Literal['fast', 'accurate'] = 'accurate',
+    video_encoder: str | None = None,
+    video_encoder_args: dict[str, Any] | None = None,
 ) -> pxt.Video | None:
     """
     Extract a clip from a video, specified by `start_time` and either `end_time` or `duration` (in seconds).
@@ -323,6 +330,14 @@ def clip(
         start_time: Start time in seconds
         end_time: End time in seconds
         duration: Duration of the clip in seconds
+        mode:
+
+            - `'fast'`: avoids re-encoding but starts the clip at the nearest keyframes and as a result, the clip
+                duration will be slightly longer than requested
+            - `'accurate'`: extracts a frame-accurate clip, but requires re-encoding
+        video_encoder: Video encoder to use. If not specified, uses the default encoder for the current platform.
+            Only available for `mode='accurate'`.
+        video_encoder_args: Additional arguments to pass to the video encoder. Only available for `mode='accurate'`.
 
     Returns:
         New video containing only the specified time range or None if start_time is beyond the end of the video.
@@ -336,6 +351,11 @@ def clip(
         raise pxt.Error(f'duration must be positive, got {duration}')
     if end_time is not None and duration is not None:
         raise pxt.Error('end_time and duration cannot both be specified')
+    if mode == 'fast':
+        if video_encoder is not None:
+            raise pxt.Error("video_encoder is not supported for mode='fast'")
+        if video_encoder_args is not None:
+            raise pxt.Error("video_encoder_args is not supported for mode='fast'")
 
     video_duration = av_utils.get_video_duration(video)
     if video_duration is not None and start_time > video_duration:
@@ -345,7 +365,15 @@ def clip(
 
     if end_time is not None:
         duration = end_time - start_time
-    cmd = av_utils.ffmpeg_clip_cmd(str(video), output_path, start_time, duration)
+    cmd = av_utils.ffmpeg_clip_cmd(
+        str(video),
+        output_path,
+        start_time,
+        duration,
+        fast=(mode == 'fast'),
+        video_encoder=video_encoder,
+        video_encoder_args=video_encoder_args,
+    )
 
     try:
         result = subprocess.run(cmd, capture_output=True, text=True, check=True)
@@ -364,7 +392,7 @@ def segment_video(
     *,
     duration: float | None = None,
     segment_times: list[float] | None = None,
-    mode: Literal['fast', 'accurate'] = 'fast',
+    mode: Literal['fast', 'accurate'] = 'accurate',
     video_encoder: str | None = None,
     video_encoder_args: dict[str, Any] | None = None,
 ) -> list[str]:
@@ -400,15 +428,14 @@ def segment_video(
     Examples:
         Split a video at 1 minute intervals using fast mode:
 
-        >>> tbl.select(segment_paths=tbl.video.segment_video(duration=60)).collect()
+        >>> tbl.select(segment_paths=tbl.video.segment_video(duration=60, mode='fast')).collect()
 
-        Split video into exact 10-second segments with accurate mode, using the libx264 encoder with a CRF of 23 and
-        slow preset (for smaller output files):
+        Split video into exact 10-second segments with default accurate mode, using the libx264 encoder with a CRF of 23
+        and slow preset (for smaller output files):
 
         >>> tbl.select(
         ...     segment_paths=tbl.video.segment_video(
         ...         duration=10,
-        ...         mode='accurate',
         ...         video_encoder='libx264',
         ...         video_encoder_args={'crf': 23, 'preset': 'slow'}
         ...     )

pixeltable/globals.py

@@ -487,12 +487,28 @@ def get_table(path: str, if_not_exists: Literal['error', 'ignore'] = 'error') ->
     return tbl
 
 
-def move(path: str, new_path: str) -> None:
+def move(
+    path: str,
+    new_path: str,
+    *,
+    if_exists: Literal['error', 'ignore'] = 'error',
+    if_not_exists: Literal['error', 'ignore'] = 'error',
+) -> None:
     """Move a schema object to a new directory and/or rename a schema object.
 
     Args:
         path: absolute path to the existing schema object.
         new_path: absolute new path for the schema object.
+        if_exists: Directive regarding how to handle if a schema object already exists at the new path.
+            Must be one of the following:
+
+            - `'error'`: raise an error
+            - `'ignore'`: do nothing and return
+        if_not_exists: Directive regarding how to handle if the source path does not exist.
+            Must be one of the following:
+
+            - `'error'`: raise an error
+            - `'ignore'`: do nothing and return
 
     Raises:
         Error: If path does not exist or new_path already exists.
@@ -506,13 +522,16 @@ def move(path: str, new_path: str) -> None:
 
         >>>> pxt.move('dir1.my_table', 'dir1.new_name')
     """
+    if_exists_ = catalog.IfExistsParam.validated(if_exists, 'if_exists')
+    if if_exists_ not in (catalog.IfExistsParam.ERROR, catalog.IfExistsParam.IGNORE):
+        raise excs.Error("`if_exists` must be one of 'error' or 'ignore'")
+    if_not_exists_ = catalog.IfNotExistsParam.validated(if_not_exists, 'if_not_exists')
     if path == new_path:
         raise excs.Error('move(): source and destination cannot be identical')
     path_obj, new_path_obj = catalog.Path.parse(path), catalog.Path.parse(new_path)
     if path_obj.is_ancestor(new_path_obj):
         raise excs.Error(f'move(): cannot move {path!r} into its own subdirectory')
-    cat = Catalog.get()
-    cat.move(path_obj, new_path_obj)
+    Catalog.get().move(path_obj, new_path_obj, if_exists_, if_not_exists_)
 
 
 def drop_table(
@@ -660,7 +679,7 @@ def _list_tables(dir_path: str = '', recursive: bool = True, allow_system_paths:
 
 
 def create_dir(
-    path: str, if_exists: Literal['error', 'ignore', 'replace', 'replace_force'] = 'error', parents: bool = False
+    path: str, *, if_exists: Literal['error', 'ignore', 'replace', 'replace_force'] = 'error', parents: bool = False
 ) -> Optional[catalog.Dir]:
     """Create a directory.
 

pixeltable/iterators/document.py

@@ -2,7 +2,7 @@ import dataclasses
 import enum
 import io
 import logging
-from typing import Any, ClassVar, Iterable, Iterator, Optional
+from typing import Any, ClassVar, Iterable, Iterator, Literal
 
 import fitz  # type: ignore[import-untyped]
 import ftfy
@@ -11,7 +11,7 @@ from bs4.element import NavigableString, Tag
 
 from pixeltable.env import Env
 from pixeltable.exceptions import Error
-from pixeltable.type_system import BoolType, ColumnType, DocumentType, ImageType, IntType, JsonType, StringType
+from pixeltable.type_system import ColumnType, DocumentType, ImageType, IntType, JsonType, StringType
 from pixeltable.utils.documents import get_document_handle
 
 from .base import ComponentIterator
@@ -19,6 +19,11 @@ from .base import ComponentIterator
 _logger = logging.getLogger('pixeltable')
 
 
+class Element(enum.Enum):
+    TEXT = 1
+    IMAGE = 2
+
+
 class ChunkMetadata(enum.Enum):
     TITLE = 1
     HEADING = 2
@@ -41,28 +46,28 @@ class DocumentSectionMetadata:
     """Metadata for a subsection of a document (ie, a structural element like a heading or paragraph)"""
 
     # html and markdown metadata
-    sourceline: Optional[int] = None
+    sourceline: int | None = None
     # the stack of headings up to the most recently observed one;
     # eg, if the most recent one was an h2, 'headings' would contain keys 1 and 2, but nothing below that
-    heading: Optional[dict[str, str]] = None
+    heading: dict[str, str] | None = None
 
     # pdf-specific metadata
-    page: Optional[int] = None
+    page: int | None = None
     # bounding box as an {x1, y1, x2, y2} dictionary
-    bounding_box: Optional[dict[str, float]] = None
+    bounding_box: dict[str, float] | None = None
 
 
 @dataclasses.dataclass
 class DocumentSection:
     """A single document chunk, according to some of the splitting criteria"""
 
-    text: Optional[str]
-    metadata: Optional[DocumentSectionMetadata]
-    image: Optional[PIL.Image.Image] = None
+    text: str | None = None
+    image: PIL.Image.Image | None = None
+    metadata: DocumentSectionMetadata | None = None
 
 
 def _parse_separators(separators: str) -> list[Separator]:
-    ret = []
+    ret: list[Separator] = []
     for s in separators.split(','):
         clean_s = s.strip().upper()
         if not clean_s:
@@ -76,7 +81,7 @@ def _parse_separators(separators: str) -> list[Separator]:
 
 
 def _parse_metadata(metadata: str) -> list[ChunkMetadata]:
-    ret = []
+    ret: list[ChunkMetadata] = []
     for m in metadata.split(','):
         clean_m = m.strip().upper()
         if not clean_m:
@@ -89,6 +94,18 @@ def _parse_metadata(metadata: str) -> list[ChunkMetadata]:
     return ret
 
 
+def _parse_elements(elements: list[Literal['text', 'image']]) -> list[Element]:
+    result: list[Element] = []
+    for e in elements:
+        clean_e = e.strip().upper()
+        if clean_e not in Element.__members__:
+            raise Error(f'Invalid element: `{e}`. Valid elements are: {", ".join(Element.__members__).lower()}')
+        result.append(Element[clean_e])
+    if len(result) == 0:
+        raise Error('elements cannot be empty')
+    return result
+
+
 _HTML_HEADINGS = {'h1', 'h2', 'h3', 'h4', 'h5', 'h6'}
 
 
@@ -106,11 +123,16 @@ class DocumentSplitter(ComponentIterator):
         separators: separators to use to chunk the document. Options are:
              `'heading'`, `'paragraph'`, `'sentence'`, `'token_limit'`, `'char_limit'`, `'page'`.
              This may be a comma-separated string, e.g., `'heading,token_limit'`.
+        elements: list of elements to extract from the document. Options are:
+            `'text'`, `'image'`. Defaults to `['text']` if not specified. The `'image'` element is only supported
+            for the `'page'` separator on PDF documents.
         limit: the maximum number of tokens or characters in each chunk, if `'token_limit'`
              or `'char_limit'` is specified.
         metadata: additional metadata fields to include in the output. Options are:
              `'title'`, `'heading'` (HTML and Markdown), `'sourceline'` (HTML), `'page'` (PDF), `'bounding_box'`
              (PDF). The input may be a comma-separated string, e.g., `'title,heading,sourceline'`.
+        image_dpi: DPI to use when extracting images from PDFs. Defaults to 300.
+        image_format: format to use when extracting images from PDFs. Defaults to 'png'.
     """
 
     METADATA_COLUMN_TYPES: ClassVar[dict[ChunkMetadata, ColumnType]] = {
@@ -121,34 +143,41 @@ class DocumentSplitter(ComponentIterator):
         ChunkMetadata.BOUNDING_BOX: JsonType(nullable=True),
     }
 
+    _doc_handle: Any
+    _separators: list[Separator]
+    _elements: list[Element]
+    _metadata_fields: list[ChunkMetadata]
+    _doc_title: str
+    _limit: int
+    _skip_tags: list[str]
+    _overlap: int
+    _tiktoken_encoding: str | None
+    _tiktoken_target_model: str | None
+    _image_dpi: int
+    _image_format: str
+
+    _sections: Iterator[DocumentSection]
+
     def __init__(
         self,
         document: str,
         *,
         separators: str,
-        limit: Optional[int] = None,
-        overlap: Optional[int] = None,
+        elements: list[Literal['text', 'image']] | None = None,
+        limit: int | None = None,
+        overlap: int | None = None,
         metadata: str = '',
-        html_skip_tags: Optional[list[str]] = None,
-        tiktoken_encoding: Optional[str] = 'cl100k_base',
-        tiktoken_target_model: Optional[str] = None,
-        # (PDF-processing-only)
-        include_page_image: bool = False,
-        page_image_dpi: int = 300,
-        page_image_format: str = 'png',
+        html_skip_tags: list[str] | None = None,
+        tiktoken_encoding: str | None = 'cl100k_base',
+        tiktoken_target_model: str | None = None,
+        image_dpi: int = 300,
+        image_format: str = 'png',
     ):
         if html_skip_tags is None:
             html_skip_tags = ['nav']
         self._doc_handle = get_document_handle(document)
+        self._elements = _parse_elements(elements.copy()) if elements is not None else [Element.TEXT]
         assert self._doc_handle is not None
-        # calling the output_schema method to validate the input arguments
-        self.output_schema(
-            separators=separators,
-            metadata=metadata,
-            limit=limit,
-            overlap=overlap,
-            include_page_image=include_page_image,
-        )
         self._separators = _parse_separators(separators)
         self._metadata_fields = _parse_metadata(metadata)
         if self._doc_handle.bs_doc is not None:
@@ -164,10 +193,8 @@ class DocumentSplitter(ComponentIterator):
         self._overlap = 0 if overlap is None else overlap
         self._tiktoken_encoding = tiktoken_encoding
         self._tiktoken_target_model = tiktoken_target_model
-
-        self._include_page_image = include_page_image
-        self._page_image_dpi = page_image_dpi
-        self._page_image_format = page_image_format
+        self._image_dpi = image_dpi
+        self._image_format = image_format
 
         # set up processing pipeline
         if self._doc_handle.format == DocumentType.DocumentFormat.HTML:
@@ -197,23 +224,28 @@ class DocumentSplitter(ComponentIterator):
         return {
             'document': DocumentType(nullable=False),
             'separators': StringType(nullable=False),
+            'elements': JsonType(nullable=False),
             'metadata': StringType(nullable=False),
             'limit': IntType(nullable=True),
             'overlap': IntType(nullable=True),
             'skip_tags': StringType(nullable=True),
             'tiktoken_encoding': StringType(nullable=True),
             'tiktoken_target_model': StringType(nullable=True),
-            # PDF options must be declared so validation accepts them:
-            'include_page_image': BoolType(nullable=True),
-            'page_image_dpi': IntType(nullable=True),
-            'page_image_format': StringType(nullable=True),
+            'image_dpi': IntType(nullable=True),
+            'image_format': StringType(nullable=True),
         }
 
     @classmethod
     def output_schema(cls, *args: Any, **kwargs: Any) -> tuple[dict[str, ColumnType], list[str]]:
-        schema: dict[str, ColumnType] = {'text': StringType()}
-        md_fields = _parse_metadata(kwargs['metadata']) if 'metadata' in kwargs else []
-
+        schema: dict[str, ColumnType] = {}
+        elements = _parse_elements(kwargs.get('elements', ['text']))
+        for element in elements:
+            if element == Element.TEXT:
+                schema['text'] = StringType(nullable=False)
+            elif element == Element.IMAGE:
+                schema['image'] = ImageType(nullable=False)
+
+        md_fields = _parse_metadata(kwargs.get('metadata', ''))
         for md_field in md_fields:
             schema[md_field.name.lower()] = cls.METADATA_COLUMN_TYPES[md_field]
 
@@ -223,6 +255,8 @@ class DocumentSplitter(ComponentIterator):
         limit = kwargs.get('limit')
         overlap = kwargs.get('overlap')
 
+        if Element.IMAGE in elements and separators != [Separator.PAGE]:
+            raise Error('Image elements are only supported for the "page" separator on PDF documents')
         if limit is not None or overlap is not None:
             if Separator.TOKEN_LIMIT not in separators and Separator.CHAR_LIMIT not in separators:
                 raise Error('limit/overlap requires the "token_limit" or "char_limit" separator')
@@ -236,23 +270,25 @@ class DocumentSplitter(ComponentIterator):
             if kwargs.get('limit') is None:
                 raise Error('limit is required with "token_limit"/"char_limit" separators')
 
-        # check dependencies at the end
         if Separator.SENTENCE in separators:
             _ = Env.get().spacy_nlp
         if Separator.TOKEN_LIMIT in separators:
             Env.get().require_package('tiktoken')
 
-        if kwargs.get('include_page_image'):
-            schema['image'] = ImageType(nullable=True)
-
         return schema, []
 
     def __next__(self) -> dict[str, Any]:
         while True:
             section = next(self._sections)
-            if section.text is None:
+            if section.text is None and section.image is None:
                 continue
-            result: dict[str, Any] = {'text': section.text}
+            result: dict[str, Any] = {}
+            for element in self._elements:
+                if element == Element.TEXT:
+                    result['text'] = section.text
+                elif element == Element.IMAGE:
+                    result['image'] = section.image
+
             for md_field in self._metadata_fields:
                 if md_field == ChunkMetadata.TITLE:
                     result[md_field.name.lower()] = self._doc_title
@@ -265,10 +301,6 @@ class DocumentSplitter(ComponentIterator):
                 elif md_field == ChunkMetadata.BOUNDING_BOX:
                     result[md_field.name.lower()] = section.metadata.bounding_box
 
-            # FIX: only include image if schema supports it
-            if self._include_page_image:
-                result['image'] = section.image
-
             return result
 
     def _html_sections(self) -> Iterator[DocumentSection]:
@@ -411,11 +443,10 @@ class DocumentSplitter(ComponentIterator):
             return txt
 
         for page_idx, page in enumerate(doc.pages()):
-            # render once per page if requested
-            page_image = None
-            if self._include_page_image:
-                pix = page.get_pixmap(dpi=self._page_image_dpi)  # ← single render
-                page_image = PIL.Image.open(io.BytesIO(pix.tobytes(self._page_image_format)))
+            img: PIL.Image.Image | None = None
+            if Element.IMAGE in self._elements:
+                pix = page.get_pixmap(dpi=self._image_dpi)
+                img = PIL.Image.open(io.BytesIO(pix.tobytes(self._image_format)))
 
             for block in page.get_text('blocks'):
                 x1, y1, x2, y2, text, *_ = block
@@ -423,14 +454,14 @@ class DocumentSplitter(ComponentIterator):
                 if accumulated_text and emit_on_paragraph:
                     bbox = {'x1': x1, 'y1': y1, 'x2': x2, 'y2': y2}
                     md = DocumentSectionMetadata(page=page_idx, bounding_box=bbox)
-                    yield DocumentSection(text=_emit_text(), metadata=md, image=page_image)
+                    yield DocumentSection(text=_emit_text(), metadata=md)
 
             if accumulated_text and emit_on_page and not emit_on_paragraph:
                 md = DocumentSectionMetadata(page=page_idx)
-                yield DocumentSection(text=_emit_text(), metadata=md, image=page_image)
+                yield DocumentSection(text=_emit_text(), image=img, metadata=md)
 
         if accumulated_text and not emit_on_page:
-            yield DocumentSection(text=_emit_text(), metadata=DocumentSectionMetadata(), image=None)
+            yield DocumentSection(text=_emit_text(), metadata=DocumentSectionMetadata())
 
     def _txt_sections(self) -> Iterator[DocumentSection]:
         """Create DocumentSections for text files.

pixeltable/iterators/video.py

@@ -251,7 +251,8 @@ class VideoSplitter(ComponentIterator):
 
     # Input parameters
     video_path: Path
-    segment_duration: float
+    segment_duration: float | None
+    segment_times: list[float] | None
     overlap: float
     min_segment_duration: float
     video_encoder: str | None
@@ -268,25 +269,31 @@ class VideoSplitter(ComponentIterator):
         self,
         video: str,
         *,
-        duration: float,
-        overlap: float = 0.0,
-        min_segment_duration: float = 0.0,
-        mode: Literal['fast', 'accurate'] = 'fast',
+        duration: float | None = None,
+        overlap: float | None = None,
+        min_segment_duration: float | None = None,
+        segment_times: list[float] | None = None,
+        mode: Literal['fast', 'accurate'] = 'accurate',
         video_encoder: str | None = None,
         video_encoder_args: dict[str, Any] | None = None,
     ):
         Env.get().require_binary('ffmpeg')
-        assert duration > 0.0
-        assert duration >= min_segment_duration
-        assert overlap < duration
+        assert (duration is not None) != (segment_times is not None)
+        if segment_times is not None:
+            assert len(segment_times) > 0
+        if duration is not None:
+            assert duration > 0.0
+            assert duration >= min_segment_duration
+            assert overlap is None or overlap < duration
 
         video_path = Path(video)
         assert video_path.exists() and video_path.is_file()
 
         self.video_path = video_path
         self.segment_duration = duration
-        self.overlap = overlap
-        self.min_segment_duration = min_segment_duration
+        self.overlap = overlap if overlap is not None else 0.0
+        self.min_segment_duration = min_segment_duration if min_segment_duration is not None else 0.0
+        self.segment_times = segment_times
         self.video_encoder = video_encoder
         self.video_encoder_args = video_encoder_args
 
@@ -304,6 +311,7 @@ class VideoSplitter(ComponentIterator):
             'duration': ts.FloatType(nullable=True),
             'overlap': ts.FloatType(nullable=True),
             'min_segment_duration': ts.FloatType(nullable=True),
+            'segment_times': ts.JsonType(nullable=True),
             'mode': ts.StringType(nullable=False),
             'video_encoder': ts.StringType(nullable=True),
             'video_encoder_args': ts.JsonType(nullable=True),
@@ -311,23 +319,34 @@ class VideoSplitter(ComponentIterator):
 
     @classmethod
     def output_schema(cls, *args: Any, **kwargs: Any) -> tuple[dict[str, ts.ColumnType], list[str]]:
-        param_names = ['duration', 'overlap', 'min_segment_duration']
+        param_names = ['duration', 'overlap', 'min_segment_duration', 'segment_times']
         params = dict(zip(param_names, args))
         params.update(kwargs)
 
-        segment_duration = params['duration']
-        min_segment_duration = params.get('min_segment_duration', 0.0)
-        overlap = params.get('overlap', 0.0)
+        segment_duration = params.get('duration')
+        segment_times = params.get('segment_times')
+        overlap = params.get('overlap')
+        min_segment_duration = params.get('min_segment_duration')
         mode = params.get('mode', 'fast')
 
-        if segment_duration <= 0.0:
-            raise excs.Error('duration must be a positive number')
-        if segment_duration < min_segment_duration:
-            raise excs.Error('duration must be at least min_segment_duration')
-        if mode == 'accurate' and overlap > 0:
+        if segment_duration is None and segment_times is None:
+            raise excs.Error('Must specify either duration or segment_times')
+        if segment_duration is not None and segment_times is not None:
+            raise excs.Error('duration and segment_times cannot both be specified')
+        if segment_times is not None:
+            if len(segment_times) == 0:
+                raise excs.Error('segment_times cannot be empty')
+            if overlap is not None:
+                raise excs.Error('overlap cannot be specified with segment_times')
+        if segment_duration is not None:
+            if segment_duration <= 0.0:
+                raise excs.Error('duration must be a positive number')
+            if min_segment_duration is not None and segment_duration < min_segment_duration:
+                raise excs.Error('duration must be at least min_segment_duration')
+            if overlap is not None and overlap >= segment_duration:
+                raise excs.Error('overlap must be less than duration')
+        if mode == 'accurate' and overlap is not None:
             raise excs.Error("Cannot specify overlap for mode='accurate'")
-        if overlap >= segment_duration:
-            raise excs.Error('overlap must be less than duration')
         if mode == 'fast':
             if params.get('video_encoder') is not None:
                 raise excs.Error("Cannot specify video_encoder for mode='fast'")
@@ -343,13 +362,22 @@ class VideoSplitter(ComponentIterator):
         }, []
 
     def fast_iter(self) -> Iterator[dict[str, Any]]:
-        segment_path: str
+        segment_path: str = ''
         try:
             start_time = 0.0
             start_pts = 0
+            segment_idx = 0
             while True:
+                target_duration: float | None
+                if self.segment_duration is not None:
+                    target_duration = self.segment_duration
+                elif self.segment_times is not None and segment_idx < len(self.segment_times):
+                    target_duration = self.segment_times[segment_idx] - start_time
+                else:
+                    target_duration = None  # the rest of the video
+
                 segment_path = str(TempStore.create_path(extension='.mp4'))
-                cmd = av_utils.ffmpeg_clip_cmd(str(self.video_path), segment_path, start_time, self.segment_duration)
+                cmd = av_utils.ffmpeg_clip_cmd(str(self.video_path), segment_path, start_time, target_duration)
                 _ = subprocess.run(cmd, capture_output=True, text=True, check=True)
 
                 # use the actual duration
@@ -373,8 +401,13 @@ class VideoSplitter(ComponentIterator):
                 start_time = segment_end - self.overlap
                 start_pts = segment_end_pts - round(self.overlap / self.video_time_base)
 
+                segment_idx += 1
+                if self.segment_times is not None and segment_idx > len(self.segment_times):
+                    # We've created all segments including the final segment after the last segment_time
+                    break
+
         except subprocess.CalledProcessError as e:
-            if Path(segment_path).exists():
+            if segment_path and Path(segment_path).exists():
                 Path(segment_path).unlink()
             error_msg = f'ffmpeg failed with return code {e.returncode}'
             if e.stderr:
@@ -389,6 +422,7 @@ class VideoSplitter(ComponentIterator):
             str(self.video_path),
             output_pattern,
             segment_duration=self.segment_duration,
+            segment_times=self.segment_times,
             video_encoder=self.video_encoder,
             video_encoder_args=self.video_encoder_args,
         )

pixeltable/plan.py

@@ -93,18 +93,13 @@ class SampleClause:
     seed: Optional[int]
     stratify_exprs: Optional[list[exprs.Expr]]
 
-    # This seed value is used if one is not supplied
-    DEFAULT_SEED = 0
-
     # The version of the hashing algorithm used for ordering and fractional sampling.
     CURRENT_VERSION = 1
 
     def __post_init__(self) -> None:
-        """If no version was provided, provide the default version"""
+        # If no version was provided, provide the default version
         if self.version is None:
             self.version = self.CURRENT_VERSION
-        if self.seed is None:
-            self.seed = self.DEFAULT_SEED
 
     @property
     def is_stratified(self) -> bool:
@@ -1006,6 +1001,7 @@ class Planner:
             analyzer.window_fn_calls
         )
         ctx = exec.ExecContext(row_builder)
+
         combined_ordering = cls._create_combined_ordering(analyzer, verify_agg=is_python_agg)
         cls._verify_join_clauses(analyzer)