Issue #747 add chunked downloading

Author: dsamaey

openeo/rest/_testing.py

@@ -103,6 +103,10 @@ def __init__(
         requests_mock.delete(
             re.compile(connection.build_url(r"/jobs/(job-\d+)/results$")), json=self._handle_delete_job_results
         )
+        requests_mock.head(
+            re.compile(connection.build_url("/jobs/(.*?)/results/result.data$")),
+            headers={"Content-Length": "666"}
+        )
         requests_mock.get(
             re.compile(connection.build_url("/jobs/(.*?)/results/result.data$")),
             content=self._handle_get_job_result_asset,

openeo/rest/job.py

@@ -5,10 +5,13 @@
 import logging
 import time
 import typing
+import urllib.error
 from pathlib import Path
 from typing import Dict, List, Optional, Union
+from urllib.error import HTTPError
 
 import requests
+import shutil
 
 from openeo.internal.documentation import openeo_endpoint
 from openeo.internal.jupyter import (
@@ -32,6 +35,8 @@
 from openeo.rest.models.logs import LogEntry, log_level_name, normalize_log_level
 from openeo.util import ensure_dir
 
+MAX_RETRIES_DOWNLOAD = 3
+
 if typing.TYPE_CHECKING:
     # Imports for type checking only (circular import issue at runtime).
     from openeo.rest.connection import Connection
@@ -402,12 +407,40 @@ def download(
             target = target / self.name
         ensure_dir(target.parent)
         logger.info("Downloading Job result asset {n!r} from {h!s} to {t!s}".format(n=self.name, h=self.href, t=target))
-        response = self._get_response(stream=True)
-        with target.open("wb") as f:
-            for block in response.iter_content(chunk_size=chunk_size):
-                f.write(block)
+        self._download_chunked(target, chunk_size)
         return target
 
+    def _download_chunked(self, target: Path, chunk_size: int):
+        file_size = None
+        try:
+            head = requests.head(self.href, stream=True)
+            if head.ok:
+                file_size = int(head.headers['Content-Length'])
+            else:
+                head.raise_for_status()
+            with target.open('wb') as f:
+                for from_byte_index in range(0, file_size, chunk_size):
+                    to_byte_index = min(from_byte_index + chunk_size - 1, file_size - 1)
+                    tries_left = MAX_RETRIES_DOWNLOAD
+                    while tries_left > 0:
+                        try:
+                            range_headers = {"Range": f"bytes={from_byte_index}-{to_byte_index}"}
+                            with requests.get(self.href, headers=range_headers, stream=True) as r:
+                                if r.ok:
+                                    shutil.copyfileobj(r.raw, f)
+                                    break
+                                else:
+                                    r.raise_for_status()
+                        except requests.exceptions.HTTPError as error:
+                            tries_left -= 1
+                            if tries_left < 1:
+                                raise error
+                            else:
+                                logger.warning(f"Failed to retrieve chunk {from_byte_index}-{to_byte_index} from {self.href} (status {error.response.status_code}) - retrying")
+                                continue
+        except requests.exceptions.HTTPError as http_error:
+            raise OpenEoApiPlainError(message=f"Failed to download {self.href}", http_status_code=http_error.response.status_code, error_message=http_error.response.text)
+
     def _get_response(self, stream=True) -> requests.Response:
         return self.job.connection.get(self.href, stream=stream)
 
@@ -501,7 +534,7 @@ def get_asset(self, name: str = None) -> ResultAsset:
                     "No asset {n!r} in: {a}".format(n=name, a=[a.name for a in assets])
                 )
 
-    def download_file(self, target: Union[Path, str] = None, name: str = None) -> Path:
+    def download_file(self, target: Union[Path, str] = None, name: str = None, chunk_size=DEFAULT_DOWNLOAD_CHUNK_SIZE) -> Path:
         """
         Download single asset. Can be used when there is only one asset in the
         :py:class:`JobResults`, or when the desired asset name is given explicitly.
@@ -513,12 +546,12 @@ def download_file(self, target: Union[Path, str] = None, name: str = None) -> Pa
         :return: path of downloaded asset
         """
         try:
-            return self.get_asset(name=name).download(target=target)
+            return self.get_asset(name=name).download(target=target, chunk_size=chunk_size)
         except MultipleAssetException:
             raise OpenEoClientException(
                 "Can not use `download_file` with multiple assets. Use `download_files` instead.")
 
-    def download_files(self, target: Union[Path, str] = None, include_stac_metadata: bool = True) -> List[Path]:
+    def download_files(self, target: Union[Path, str] = None, include_stac_metadata: bool = True, chunk_size=DEFAULT_DOWNLOAD_CHUNK_SIZE) -> List[Path]:
         """
         Download all assets to given folder.
 
@@ -531,7 +564,7 @@ def download_files(self, target: Union[Path, str] = None, include_stac_metadata:
             raise OpenEoClientException(f"Target argument {target} exists but isn't a folder.")
         ensure_dir(target)
 
-        downloaded = [a.download(target) for a in self.get_assets()]
+        downloaded = [a.download(target, chunk_size=chunk_size) for a in self.get_assets()]
 
         if include_stac_metadata:
             # TODO #184: convention for metadata file name?

tests/rest/test_job.py

@@ -1,3 +1,4 @@
+import inspect
 import itertools
 import json
 import logging
@@ -11,16 +12,14 @@
 
 import openeo
 import openeo.rest.job
-from openeo.rest import JobFailedException, OpenEoApiPlainError, OpenEoClientException
+from openeo.rest import JobFailedException, OpenEoApiPlainError, OpenEoClientException, DEFAULT_DOWNLOAD_CHUNK_SIZE
 from openeo.rest.job import BatchJob, ResultAsset
 from openeo.rest.models.general import Link
 from openeo.rest.models.logs import LogEntry
 
 API_URL = "https://oeo.test"
 
-TIFF_CONTENT = b'T1f7D6t6l0l' * 1000
-
-
+TIFF_CONTENT = b'T1f7D6t6l0l' * 10000
 
 @pytest.fixture
 def con100(requests_mock):
@@ -74,7 +73,7 @@ def test_execute_batch(con100, requests_mock, tmpdir):
             }
         },
     )
-    requests_mock.get(API_URL + "/jobs/f00ba5/files/output.tiff", text="tiffdata")
+    _mock_get_head_content(requests_mock, API_URL + "/jobs/f00ba5/files/output.tiff", "tiffdata")
     requests_mock.get(API_URL + "/jobs/f00ba5/logs", json={'logs': []})
 
     path = tmpdir.join("tmp.tiff")
@@ -231,7 +230,8 @@ def test_execute_batch_with_soft_errors(con100, requests_mock, tmpdir, error_res
             }
         },
     )
-    requests_mock.get(API_URL + "/jobs/f00ba5/files/output.tiff", text="tiffdata")
+    _mock_get_head_content(requests_mock, API_URL + "/jobs/f00ba5/files/output.tiff", "tiffdata")
+    # requests_mock.get(API_URL + "/jobs/f00ba5/files/output.tiff", text="tiffdata")
     requests_mock.get(API_URL + "/jobs/f00ba5/logs", json={'logs': []})
 
     path = tmpdir.join("tmp.tiff")
@@ -536,10 +536,28 @@ def job_with_1_asset(con100, requests_mock, tmp_path) -> BatchJob:
     requests_mock.get(API_URL + "/jobs/jj1/results", json={"assets": {
         "1.tiff": {"href": API_URL + "/dl/jjr1.tiff", "type": "image/tiff; application=geotiff"},
     }})
+    requests_mock.head(API_URL + "/dl/jjr1.tiff", headers={"Content-Length": f"{len(TIFF_CONTENT)}"})
     requests_mock.get(API_URL + "/dl/jjr1.tiff", content=TIFF_CONTENT)
+
     job = BatchJob("jj1", connection=con100)
     return job
 
+@pytest.fixture
+def job_with_chunked_asset(con100, requests_mock, tmp_path) -> BatchJob:
+    requests_mock.get(API_URL + "/jobs/jj1/results", json={"assets": {
+        "1.tiff": {"href": API_URL + "/dl/jjr1.tiff", "type": "image/tiff; application=geotiff"},
+    }})
+    requests_mock.head(API_URL + "/dl/jjr1.tiff", headers={"Content-Length": f"{len(TIFF_CONTENT)}"})
+
+    chunk_size = 1000
+    for r in range(0, len(TIFF_CONTENT), chunk_size):
+        from_bytes = r
+        to_bytes = min(r + chunk_size, len(TIFF_CONTENT)) - 1
+        # fail the 1st time, serve the content chunk the 2nd time
+        requests_mock.get(API_URL + "/dl/jjr1.tiff", request_headers={"Range": f"bytes={from_bytes}-{to_bytes}"},
+                     response_list = [{"status_code": 500, "text": "Server error"},     {"content": TIFF_CONTENT[from_bytes:to_bytes+1]}])
+    job = BatchJob("jj1", connection=con100)
+    return job
 
 @pytest.fixture
 def job_with_2_assets(con100, requests_mock, tmp_path) -> BatchJob:
@@ -551,8 +569,11 @@ def job_with_2_assets(con100, requests_mock, tmp_path) -> BatchJob:
             "2.tiff": {"href": API_URL + "/dl/jjr2.tiff", "type": "image/tiff; application=geotiff"},
         }
     })
+    requests_mock.head(API_URL + "/dl/jjr1.tiff", headers={"Content-Length": f"{len(TIFF_CONTENT)}"})
     requests_mock.get(API_URL + "/dl/jjr1.tiff", content=TIFF_CONTENT)
+    requests_mock.head(API_URL + "/dl/jjr2.tiff", headers={"Content-Length": f"{len(TIFF_CONTENT)}"})
     requests_mock.get(API_URL + "/dl/jjr2.tiff", content=TIFF_CONTENT)
+
     job = BatchJob("jj2", connection=con100)
     return job
 
@@ -574,6 +595,13 @@ def test_get_results_download_file(job_with_1_asset: BatchJob, tmp_path):
     with target.open("rb") as f:
         assert f.read() == TIFF_CONTENT
 
+def test_get_results_download_chunked_file(job_with_chunked_asset: BatchJob, tmp_path):
+    job = job_with_chunked_asset
+    target = tmp_path / "result.tiff"
+    res = job.get_results().download_file(target, chunk_size=1000)
+    assert res == target
+    with target.open("rb") as f:
+        assert f.read() == TIFF_CONTENT
 
 def test_download_result_folder(job_with_1_asset: BatchJob, tmp_path):
     job = job_with_1_asset
@@ -714,7 +742,7 @@ def test_get_results_download_files_include_stac_metadata(
 
 def test_result_asset_download_file(con100, requests_mock, tmp_path):
     href = API_URL + "/dl/jjr1.tiff"
-    requests_mock.get(href, content=TIFF_CONTENT)
+    _mock_get_head_content(requests_mock, href, TIFF_CONTENT)
 
     job = BatchJob("jj", connection=con100)
     asset = ResultAsset(job, name="1.tiff", href=href, metadata={'type': 'image/tiff; application=geotiff'})
@@ -729,6 +757,7 @@ def test_result_asset_download_file(con100, requests_mock, tmp_path):
 
 def test_result_asset_download_file_error(con100, requests_mock, tmp_path):
     href = API_URL + "/dl/jjr1.tiff"
+    requests_mock.head(href, status_code=500, text="Nope!")
     requests_mock.get(href, status_code=500, text="Nope!")
 
     job = BatchJob("jj", connection=con100)
@@ -743,7 +772,7 @@ def test_result_asset_download_file_error(con100, requests_mock, tmp_path):
 
 def test_result_asset_download_folder(con100, requests_mock, tmp_path):
     href = API_URL + "/dl/jjr1.tiff"
-    requests_mock.get(href, content=TIFF_CONTENT)
+    _mock_get_head_content(requests_mock, href, TIFF_CONTENT)
 
     job = BatchJob("jj", connection=con100)
     asset = ResultAsset(job, name="1.tiff", href=href, metadata={"type": "image/tiff; application=geotiff"})
@@ -770,7 +799,7 @@ def test_result_asset_load_json(con100, requests_mock):
 
 def test_result_asset_load_bytes(con100, requests_mock):
     href = API_URL + "/dl/jjr1.tiff"
-    requests_mock.get(href, content=TIFF_CONTENT)
+    _mock_get_head_content(requests_mock, href, TIFF_CONTENT)
 
     job = BatchJob("jj", connection=con100)
     asset = ResultAsset(job, name="out.tiff", href=href, metadata={"type": "image/tiff; application=geotiff"})
@@ -797,6 +826,7 @@ def download_tiff(request, context):
         return TIFF_CONTENT
 
     requests_mock.get(API_URL + "/jobs/jj1/results", json=get_results)
+    requests_mock.head("https://evilcorp.test/dl/jjr1.tiff", headers={"Content-Length": "666"})
     requests_mock.get("https://evilcorp.test/dl/jjr1.tiff", content=download_tiff)
 
     con100.authenticate_basic("john", "j0hn")
@@ -880,3 +910,15 @@ def get_jobs(request, context):
     assert jobs.links == [Link(rel="next", href="https://oeo.test/jobs?limit=2&offset=2")]
     assert jobs.ext_federation_missing() == ["oeob"]
     assert "Partial job listing: missing federation components: ['oeob']." in caplog.text
+
+
+def _mock_get_head_content(requests_mock, url: str, content):
+    if callable(content):
+        requests_mock.head(url, headers={"Content-Length": "666"})
+        requests_mock.get(url, content=content)
+    elif type(content) == str:
+        requests_mock.head(url, headers={"Content-Length": f"{len(content)}"})
+        requests_mock.get(url, text=content)
+    else:
+        requests_mock.head(url, headers={"Content-Length": f"{len(content)}"})
+        requests_mock.get(url, content=content)