HDXDSYS-2931 Make downloads resumable

documentation/index.md

@@ -205,7 +205,15 @@ Other useful functions:
     # if no folder supplied
     # path = Download.get_path_for_url(url, folder)
 
-For more detail and additional functions, check the API documentation.
+### Resumable downloads
+
+Pass `resume=True` and a fixed `path` to continue an interrupted download using
+HTTP Range requests. If the server returns 200 instead of 206, the file is
+downloaded from scratch. Add `retries` to automatically retry mid-stream
+failures; each retry continues from the current end of the partial file.
+`retries` has no effect without `resume=True`.
+
+    path = downloader.download_file(url, path=mypath, resume=True, retries=3)
 
 ## Retrieving files
 
@@ -261,6 +269,8 @@ Examples:
         retriever = Retrieve(downloader, fallback_dir, saved_dir, temp_dir, save=False, use_saved=True)
         data = retriever.download_json(url, filename, logstr="test json", fallback=False, log_level=logging.DEBUG)
 
+`Retrieve.download_file` also accepts `resume=True` and `retries=N` — see [Resumable downloads](#resumable-downloads).
+
 ## Date utilities
 
 There are utilities to parse dates. By default, *no timezone information will be parsed

pyproject.toml

@@ -70,6 +70,7 @@ dev = [
     "pytest",
     "pytest-cov",
     "pytest-loguru",
+    "pytest-mock",
     "pre-commit",
     "ruff==0.14.13",
 ]

src/hdx/utilities/downloader.py

@@ -250,21 +250,24 @@ def hash_stream(self, url: Path | str) -> str:
                 f"Download of {url} failed in retrieval of stream!" % url
             )
 
-    def stream_path(self, path: Path | str, errormsg: str) -> Path:
+    def stream_path(
+        self, path: Path | str, errormsg: str, append: bool = False
+    ) -> Path:
         """Stream file from url and store in provided path. Must call setup
         method first.
 
         Args:
             path: Path for downloaded file
             errormsg: Error message to display if there is a problem
+            append: Whether to append to an existing file. Defaults to False.
 
         Returns:
             Path of downloaded file
         """
         f = None
         try:
             path = Path(path)
-            f = path.open("wb")
+            f = path.open("ab" if append else "wb")
             for chunk in self.response.iter_content(chunk_size=10240):
                 if chunk:  # filter out keep-alive new chunks
                     f.write(chunk)
@@ -325,6 +328,8 @@ def download_file(
             path (str): Full path to use for downloaded file instead of folder and filename.
             overwrite (bool): Whether to overwrite existing file. Defaults to False.
             keep (bool): Whether to keep already downloaded file. Defaults to False.
+            resume (bool): Whether to resume a partial download using Range requests where the server supports it. Defaults to False.
+            retries (int): Number of times to retry a mid-stream failure. Only effective when resume=True so that each retry can continue from the partial file. Defaults to 0.
             post (bool): Whether to use POST instead of GET. Defaults to False.
             parameters (dict): Parameters to pass. Defaults to None.
             timeout (float): Timeout for connecting to URL. Defaults to None (no timeout).
@@ -340,25 +345,84 @@ def download_file(
         path = kwargs.get("path")
         overwrite = kwargs.get("overwrite", False)
         keep = kwargs.get("keep", False)
+        resume = kwargs.get("resume", False)
+        retries = kwargs.get("retries", 0)
         try:
-            path = get_path_for_url(url, folder, filename, path, overwrite, keep)
+            # When resuming, skip uniqueness renaming so we get back the exact target path
+            path = get_path_for_url(
+                url, folder, filename, path, overwrite, keep or resume
+            )
         except ValueError as ex:
             raise DownloadError(ex) from ex
-        if keep and exists(path):
+        if keep and not resume and exists(path):
             return path
-        self.setup(
-            url,
-            stream=True,
-            post=kwargs.get("post", False),
-            parameters=kwargs.get("parameters"),
-            timeout=kwargs.get("timeout"),
-            headers=kwargs.get("headers"),
-            encoding=kwargs.get("encoding"),
-            json_string=kwargs.get("json_string", False),
-        )
-        return self.stream_path(
-            path, f"Download of {url} failed in retrieval of stream!"
-        )
+        errormsg = f"Download of {url} failed in retrieval of stream!"
+        setup_headers = dict(kwargs.get("headers") or {})
+        append = False
+        if resume and path.exists() and path.stat().st_size > 0:
+            setup_headers["Range"] = f"bytes={path.stat().st_size}-"
+            # Range bytes must come from the raw stream: partial gzip streams
+            # lack the header and cannot be decoded, so request uncompressed.
+            setup_headers.setdefault("Accept-Encoding", "identity")
+            try:
+                self.setup(
+                    url,
+                    stream=True,
+                    post=kwargs.get("post", False),
+                    parameters=kwargs.get("parameters"),
+                    timeout=kwargs.get("timeout"),
+                    headers=setup_headers,
+                    encoding=kwargs.get("encoding"),
+                    json_string=kwargs.get("json_string", False),
+                )
+            except DownloadError:
+                if self.response is not None and self.response.status_code == 416:
+                    return path
+                raise
+            append = self.response.status_code == 206
+        else:
+            self.setup(
+                url,
+                stream=True,
+                post=kwargs.get("post", False),
+                parameters=kwargs.get("parameters"),
+                timeout=kwargs.get("timeout"),
+                headers=setup_headers or None,
+                encoding=kwargs.get("encoding"),
+                json_string=kwargs.get("json_string", False),
+            )
+        attempts_remaining = retries
+        while True:
+            try:
+                return self.stream_path(path, errormsg, append=append)
+            except DownloadError:
+                if not resume or attempts_remaining == 0:
+                    raise
+                attempts_remaining -= 1
+                logger.warning(
+                    "Download of %s interrupted, retrying (attempt %d of %d)...",
+                    url,
+                    retries - attempts_remaining,
+                    retries,
+                )
+                setup_headers["Range"] = f"bytes={path.stat().st_size}-"
+                setup_headers.setdefault("Accept-Encoding", "identity")
+                try:
+                    self.setup(
+                        url,
+                        stream=True,
+                        post=kwargs.get("post", False),
+                        parameters=kwargs.get("parameters"),
+                        timeout=kwargs.get("timeout"),
+                        headers=setup_headers,
+                        encoding=kwargs.get("encoding"),
+                        json_string=kwargs.get("json_string", False),
+                    )
+                except DownloadError:
+                    if self.response is not None and self.response.status_code == 416:
+                        return path
+                    raise
+                append = self.response.status_code == 206
 
     def download(self, url: Path | str, **kwargs: Any) -> requests.Response:
         """Download url.

src/hdx/utilities/retriever.py

@@ -184,6 +184,8 @@ def download_file(
         filename: str | None = None,
         logstr: str | None = None,
         fallback: bool = False,
+        resume: bool = False,
+        retries: int = 0,
         log_level: int = None,
         **kwargs: Any,
     ) -> Path:
@@ -194,6 +196,8 @@ def download_file(
             filename: Filename of saved file. Defaults to getting from url.
             logstr: Text to use in log string to describe download. Defaults to filename.
             fallback: Whether to use static fallback if download fails. Defaults to False.
+            resume: Whether to resume a partial download using Range requests where the server supports it. Defaults to False.
+            retries: Number of times to retry a mid-stream failure. Only effective when resume=True. Defaults to 0.
             log_level: Level at which to log messages. Overrides level from constructor.
             **kwargs: Parameters to pass to download_file call
 
@@ -219,7 +223,9 @@ def download_file(
                 log_level,
                 f"Downloading {logstr} from {self.get_url_logstr(url)} into {output_path}",
             )
-            return self.downloader.download_file(url, path=output_path, **kwargs)
+            return self.downloader.download_file(
+                url, path=output_path, resume=resume, retries=retries, **kwargs
+            )
         except DownloadError:
             if not fallback:
                 raise

tests/hdx/utilities/test_downloader.py

@@ -1314,3 +1314,159 @@ def test_generate_downloaders(self, downloaders):
         downloader1 = Download.get_downloader()
         downloader2 = Download.get_downloader("test")
         assert_downloaders(downloader1, downloader2, downloaders)
+
+
+class TestResumableDownload:
+    url = "https://example.com/testfile.bin"
+
+    def _make_response(self, mocker, status_code, content=b""):
+        import requests
+
+        mock_resp = mocker.MagicMock()
+        mock_resp.status_code = status_code
+        if status_code >= 400:
+            mock_resp.raise_for_status.side_effect = requests.exceptions.HTTPError(
+                response=mock_resp
+            )
+        else:
+            mock_resp.raise_for_status.return_value = None
+        mock_resp.iter_content.return_value = iter([content] if content else [])
+        return mock_resp
+
+    def test_resume_206_appends(self, tmp_path, mocker):
+        partial = tmp_path / "file.bin"
+        partial.write_bytes(b"hello ")
+        mock_resp = self._make_response(mocker, 206, b"world")
+        with Download(user_agent="test") as downloader:
+            mocker.patch.object(downloader.session, "get", return_value=mock_resp)
+            result = downloader.download_file(self.url, path=partial, resume=True)
+        assert result == partial
+        assert partial.read_bytes() == b"hello world"
+        sent_headers = downloader.session.get.call_args.kwargs["headers"]
+        assert sent_headers["Range"] == "bytes=6-"
+
+    def test_resume_200_overwrites(self, tmp_path, mocker):
+        partial = tmp_path / "file.bin"
+        partial.write_bytes(b"partial")
+        mock_resp = self._make_response(mocker, 200, b"fullcontent")
+        with Download(user_agent="test") as downloader:
+            mocker.patch.object(downloader.session, "get", return_value=mock_resp)
+            downloader.download_file(self.url, path=partial, resume=True)
+        assert partial.read_bytes() == b"fullcontent"
+
+    def test_resume_416_returns_existing(self, tmp_path, mocker):
+        partial = tmp_path / "file.bin"
+        partial.write_bytes(b"complete")
+        mock_resp = self._make_response(mocker, 416)
+        with Download(user_agent="test") as downloader:
+            mocker.patch.object(downloader.session, "get", return_value=mock_resp)
+            result = downloader.download_file(self.url, path=partial, resume=True)
+        assert result == partial
+        assert partial.read_bytes() == b"complete"
+
+    def test_resume_no_partial_file_does_full_download(self, tmp_path, mocker):
+        path = tmp_path / "newfile.bin"
+        mock_resp = self._make_response(mocker, 200, b"fullcontent")
+        with Download(user_agent="test") as downloader:
+            mocker.patch.object(downloader.session, "get", return_value=mock_resp)
+            downloader.download_file(self.url, path=path, resume=True)
+        assert path.read_bytes() == b"fullcontent"
+        sent_headers = downloader.session.get.call_args.kwargs.get("headers")
+        assert sent_headers is None or "Range" not in sent_headers
+
+    def test_no_resume_overwrites_partial(self, tmp_path, mocker):
+        partial = tmp_path / "file.bin"
+        partial.write_bytes(b"staledata")
+        mock_resp = self._make_response(mocker, 200, b"freshcontent")
+        with Download(user_agent="test") as downloader:
+            mocker.patch.object(downloader.session, "get", return_value=mock_resp)
+            downloader.download_file(self.url, path=partial, overwrite=True)
+        assert partial.read_bytes() == b"freshcontent"
+        sent_headers = downloader.session.get.call_args.kwargs.get("headers")
+        assert sent_headers is None or "Range" not in sent_headers
+
+    def test_resume_real_file(self, tmp_path, fixturesfolder):
+        fixture = fixturesfolder / "downloader" / "resume_test.dat"
+        expected = fixture.read_bytes()
+        split_at = 40
+        partial = tmp_path / "resume_test.dat"
+        partial.write_bytes(expected[:split_at])
+        url = "https://raw.githubusercontent.com/OCHA-DAP/hdx-python-utilities/main/tests/fixtures/downloader/resume_test.dat"
+        with Download(user_agent="test") as downloader:
+            result = downloader.download_file(url, path=partial, resume=True)
+        assert result == partial
+        assert partial.read_bytes() == expected
+
+    def test_retry_succeeds_on_second_attempt(self, tmp_path, mocker):
+        partial = tmp_path / "file.bin"
+        partial.write_bytes(b"hello ")
+        first_resp = self._make_response(mocker, 206, b"wo")
+        first_resp.iter_content.side_effect = Exception("connection reset")
+        second_resp = self._make_response(mocker, 206, b"world")
+        with Download(user_agent="test") as downloader:
+            mocker.patch.object(
+                downloader.session, "get", side_effect=[first_resp, second_resp]
+            )
+            result = downloader.download_file(
+                self.url, path=partial, resume=True, retries=1
+            )
+        assert result == partial
+        assert partial.read_bytes() == b"hello world"
+        second_call_headers = downloader.session.get.call_args_list[1].kwargs["headers"]
+        assert second_call_headers["Range"] == "bytes=6-"
+
+    def test_retry_exhausted_raises(self, tmp_path, mocker):
+        partial = tmp_path / "file.bin"
+        partial.write_bytes(b"hello ")
+        failing_resp = self._make_response(mocker, 206, b"world")
+        failing_resp.iter_content.side_effect = Exception("connection reset")
+        with Download(user_agent="test") as downloader:
+            mocker.patch.object(
+                downloader.session,
+                "get",
+                return_value=failing_resp,
+            )
+            with pytest.raises(DownloadError):
+                downloader.download_file(self.url, path=partial, resume=True, retries=2)
+        assert downloader.session.get.call_count == 3
+
+    def test_retry_not_attempted_without_resume(self, tmp_path, mocker):
+        path = tmp_path / "file.bin"
+        failing_resp = self._make_response(mocker, 200, b"content")
+        failing_resp.iter_content.side_effect = Exception("connection reset")
+        with Download(user_agent="test") as downloader:
+            mocker.patch.object(downloader.session, "get", return_value=failing_resp)
+            with pytest.raises(DownloadError):
+                downloader.download_file(self.url, path=path, retries=3)
+        assert downloader.session.get.call_count == 1
+
+    def test_retry_updates_range_from_partial_bytes(self, tmp_path, mocker):
+        partial = tmp_path / "file.bin"
+        partial.write_bytes(b"hello ")
+
+        def make_partial_resp(content, fail_after):
+            resp = self._make_response(mocker, 206, b"")
+            chunks = [content[i : i + 1] for i in range(len(content))]
+
+            def iter_content_side_effect(chunk_size=1):
+                for i, chunk in enumerate(chunks):
+                    if i == fail_after:
+                        raise Exception("connection reset")
+                    yield chunk
+
+            resp.iter_content.side_effect = iter_content_side_effect
+            return resp
+
+        first_resp = make_partial_resp(b"wor", fail_after=2)
+        second_resp = self._make_response(mocker, 206, b"rld")
+        with Download(user_agent="test") as downloader:
+            mocker.patch.object(
+                downloader.session, "get", side_effect=[first_resp, second_resp]
+            )
+            result = downloader.download_file(
+                self.url, path=partial, resume=True, retries=1
+            )
+        assert result == partial
+        assert partial.read_bytes() == b"hello world"
+        second_call_headers = downloader.session.get.call_args_list[1].kwargs["headers"]
+        assert second_call_headers["Range"] == "bytes=8-"

tests/hdx/utilities/test_retriever.py

@@ -462,6 +462,27 @@ def test_download_usesaved(self, dirs, retrieverfolder, fallback_dir):
                 # Uses saved file
                 assert headers == ["header1", "header2", "header3", "header4"]
 
+    def test_resume_forwarded_to_downloader(self, dirs, fallback_dir, mocker):
+        saved_dir, temp_dir = dirs
+        downloader = mocker.MagicMock(spec=Download)
+        filename = "test.txt"
+        expected_path = temp_dir / filename
+        downloader.download_file.return_value = expected_path
+        with Retrieve(
+            downloader,
+            fallback_dir,
+            saved_dir,
+            temp_dir,
+            save=False,
+            use_saved=False,
+        ) as retriever:
+            url = "http://example.com/test.txt"
+            path = retriever.download_file(url, filename, resume=True)
+            assert path == expected_path
+            downloader.download_file.assert_called_once_with(
+                url, path=expected_path, resume=True, retries=0
+            )
+
     def test_generate_retrievers(self, downloaders, dirs, fallback_dir):
         saved_dir, temp_dir = dirs
         Retrieve.generate_retrievers(