datamarket 0.9.48__tar.gz → 0.9.50__tar.gz

{datamarket-0.9.48 → datamarket-0.9.50}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: datamarket
-Version: 0.9.48
+Version: 0.9.50
 Summary: Utilities that integrate advanced scraping knowledge into just one library.
 License: GPL-3.0-or-later
 Author: DataMarket

{datamarket-0.9.48 → datamarket-0.9.50}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "datamarket"
-version = "0.9.48"
+version = "0.9.50"
 description = "Utilities that integrate advanced scraping knowledge into just one library."
 authors = ["DataMarket <techsupport@datamarket.es>"]
 license = "GPL-3.0-or-later"

{datamarket-0.9.48 → datamarket-0.9.50}/src/datamarket/interfaces/drive.py

@@ -18,13 +18,9 @@ class DriveInterface:
         if "drive" in config:
             self.config = config["drive"]
 
-            GoogleAuth.DEFAULT_SETTINGS[
-                "client_config_file"
-            ] = f'{self.config["config_path"]}/credentials.json'
+            GoogleAuth.DEFAULT_SETTINGS["client_config_file"] = f"{self.config['config_path']}/credentials.json"
 
-            self.gauth = GoogleAuth(
-                settings_file=f'{self.config["config_path"]}/settings.yaml'
-            )
+            self.gauth = GoogleAuth(settings_file=f"{self.config['config_path']}/settings.yaml")
             self.gauth.LocalWebserverAuth()
 
             self.drive = GoogleDrive(self.gauth)
@@ -48,24 +44,78 @@ class DriveInterface:
                 logger.info(f"deleting old {filename}...")
                 drive_file.Delete(param={"supportsTeamDrives": True})
 
+    def _create_remote_dir_tree(self, base_folder_id, path_parts):
+        """
+        Ensure the nested folders described by path_parts exist under base_folder_id.
+        Returns the folder_id of the deepest folder (or base_folder_id if path_parts is empty).
+        """
+        parent_id = base_folder_id
+        for part in path_parts:
+            part = part.strip()
+            if not part:
+                continue
+
+            query = (
+                f"'{parent_id}' in parents and title = '{part}'"
+                " and mimeType = 'application/vnd.google-apps.folder' and trashed=false"
+            )
+            results = self.drive.ListFile(
+                {
+                    "q": query,
+                    "corpora": "teamDrive",
+                    "teamDriveId": self.team_id,
+                    "includeTeamDriveItems": True,
+                    "supportsTeamDrives": True,
+                }
+            ).GetList()
+
+            if results:
+                parent_id = results[0]["id"]
+            else:
+                folder_metadata = {
+                    "title": part,
+                    "mimeType": "application/vnd.google-apps.folder",
+                    "parents": [
+                        {
+                            "kind": "drive#fileLink",
+                            "teamDriveId": self.team_id,
+                            "id": parent_id,
+                        }
+                    ],
+                }
+                folder = self.drive.CreateFile(folder_metadata)
+                folder.Upload(param={"supportsTeamDrives": True})
+                parent_id = folder["id"]
+
+        return parent_id
+
     def upload_file(self, local_filename, drive_filename, folder_id):
-        self.delete_old_files(drive_filename, folder_id)
+        drive_filename = drive_filename.strip("/")
+        parts = drive_filename.split("/")
+        if len(parts) > 1:
+            *folders, filename = parts
+            target_folder_id = self._create_remote_dir_tree(folder_id, folders)
+        else:
+            filename = parts[0]
+            target_folder_id = folder_id
+
+        self.delete_old_files(filename, target_folder_id)
 
         f = self.drive.CreateFile(
             {
-                "title": drive_filename,
+                "title": filename,
                 "parents": [
                     {
                         "kind": "drive#fileLink",
                         "teamDriveId": self.team_id,
-                        "id": folder_id,
+                        "id": target_folder_id,
                     }
                 ],
             }
         )
         f.SetContentFile(local_filename)
 
-        logger.info(f"uploading {drive_filename} to folder: {folder_id}...")
+        logger.info(f"uploading {drive_filename} to folder: {target_folder_id}...")
         f.Upload(param={"supportsTeamDrives": True})
 
     def validate_file(self, filename, folder_id):

datamarket-0.9.50/src/datamarket/utils/playwright/async_api.py

@@ -0,0 +1,152 @@
+########################################################################################################################
+# IMPORTS
+
+import asyncio
+import logging
+from datetime import timedelta
+from random import randint
+from types import TracebackType
+from typing import Optional, Self
+
+# 'BdbQuit' import is removed as it's no longer used
+from camoufox.async_api import AsyncCamoufox as Camoufox
+from playwright.async_api import (
+    Browser,
+    BrowserContext,
+    Error as PlaywrightError,
+    Page,
+    TimeoutError as PlaywrightTimeoutError,
+)
+from tenacity import (
+    before_sleep_log,
+    retry,
+    retry_if_exception_type,
+    stop_after_delay,
+    wait_exponential,
+)
+
+from datamarket.interfaces.proxy import ProxyInterface
+
+
+########################################################################################################################
+# SETUP LOGGER
+
+logger = logging.getLogger(__name__)
+
+########################################################################################################################
+# ASYNC HELPER FUNCTIONS
+
+
+async def human_type(page: Page, text: str, delay: int = 100):
+    for char in text:
+        await page.keyboard.type(char, delay=randint(int(delay * 0.5), int(delay * 1.5)))
+
+
+async def human_press_key(page: Page, key: str, count: int = 1, delay: int = 100, add_sleep: bool = True) -> None:
+    """Asynchronously presses a key with a random delay, optionally sleeping between presses."""
+    for _ in range(count):
+        await page.keyboard.press(key, delay=randint(int(delay * 0.5), int(delay * 1.5)))
+        if add_sleep:
+            await asyncio.sleep(randint(int(delay * 1.5), int(delay * 2.5)) / 1000)
+
+
+########################################################################################################################
+# ASYNC CRAWLER CLASS
+
+
+class PlaywrightCrawler:
+    """An robust, proxy-enabled asynchronous Playwright crawler with captcha bypass and retry logic."""
+
+    def __init__(self, proxy_interface: ProxyInterface):
+        """
+        Initializes the async crawler with a proxy interface.
+
+        Args:
+            proxy_interface (ProxyInterface): An async-compatible object to fetch proxy credentials.
+        """
+        self.proxy_interface = proxy_interface
+        self.pw: Optional[Camoufox] = None
+        self.browser: Optional[Browser] = None
+        self.context: Optional[BrowserContext] = None
+        self.page: Optional[Page] = None
+
+    async def __aenter__(self) -> Self:
+        """Initializes the browser context when entering the `async with` statement."""
+        await self.init_context()
+        return self
+
+    async def __aexit__(
+        self,
+        exc_type: Optional[type[BaseException]],
+        exc_val: Optional[BaseException],
+        exc_tb: Optional[TracebackType],
+    ) -> None:
+        """Safely closes the browser context upon exit."""
+        if self.pw:
+            await self.pw.__aexit__(exc_type, exc_val, exc_tb)
+
+    @retry(
+        wait=wait_exponential(exp_base=2, multiplier=3, max=90),
+        stop=stop_after_delay(timedelta(minutes=10)),
+        before_sleep=before_sleep_log(logger, logging.INFO),
+        reraise=True,
+    )
+    async def init_context(self) -> Self:
+        """Initializes a new async browser instance and context with a fresh proxy."""
+        try:
+            # Correctly wrap the blocking I/O call
+            host, port, user, pwd = await asyncio.to_thread(self.proxy_interface.get_proxies, raw=True, use_auth=True)
+            proxy_url = f"http://{host}:{port}"
+            proxy_cfg = {"server": proxy_url}
+
+            if user and pwd:
+                proxy_cfg.update({"username": user, "password": pwd})
+
+            logger.info(f"Starting browser with proxy: {proxy_url}")
+            self.pw = Camoufox(headless=True, geoip=True, humanize=True, proxy=proxy_cfg)
+            self.browser = await self.pw.__aenter__()
+            self.context = await self.browser.new_context()
+            self.page = await self.context.new_page()
+        except Exception as e:
+            logger.error(f"Failed to initialize browser context: {e}")
+            if self.pw:
+                await self.pw.__aexit__(type(e), e, e.__traceback__)
+            raise
+        return self
+
+    async def restart_context(self) -> None:
+        """Closes the current browser instance and initializes a new one."""
+        logger.info("Restarting browser context...")
+        if self.pw:
+            await self.pw.__aexit__(None, None, None)
+        await self.init_context()
+
+    @retry(
+        retry=retry_if_exception_type((PlaywrightTimeoutError, PlaywrightError)),
+        wait=wait_exponential(exp_base=2, multiplier=3, max=90),
+        stop=stop_after_delay(timedelta(minutes=10)),
+        before_sleep=before_sleep_log(logger, logging.INFO),
+        reraise=True,
+    )
+    async def _goto_with_retry(self, url: str) -> Page:
+        """
+        Asynchronously navigates to a URL with retries for common Playwright errors.
+        Restarts the browser context on repeated failures.
+        """
+        if not (self.page and not self.page.is_closed()):
+            logger.warning("Page is not available or closed. Restarting context.")
+            await self.restart_context()
+
+        assert self.page is not None
+        await self.page.goto(url, timeout=30000, wait_until="domcontentloaded")
+        return self.page
+
+    async def goto(self, url: str) -> Page:
+        """
+        Ensures the browser is initialized and navigates to the given URL.
+        Public wrapper for the internal retry-enabled navigation method.
+        """
+        if not self.page:
+            logger.info("Browser context not found, initializing now...")
+            await self.init_context()
+        return await self._goto_with_retry(url)

datamarket-0.9.50/src/datamarket/utils/playwright/sync_api.py

@@ -0,0 +1,149 @@
+########################################################################################################################
+# IMPORTS
+
+import logging
+import time
+from datetime import timedelta
+from random import randint
+from types import TracebackType
+from typing import Optional, Self
+
+from camoufox import Camoufox
+from playwright.sync_api import (
+    Browser,
+    BrowserContext,
+    Error as PlaywrightError,
+    Page,
+    TimeoutError as PlaywrightTimeoutError,
+)
+from tenacity import (
+    before_sleep_log,
+    retry,
+    retry_if_exception_type,
+    stop_after_delay,
+    wait_exponential,
+)
+from datamarket.interfaces.proxy import ProxyInterface
+
+########################################################################################################################
+# SETUP LOGGER
+
+logger = logging.getLogger(__name__)
+
+
+########################################################################################################################
+# HELPER FUNCTIONS
+def human_type(page: Page, text: str, delay: int = 100):
+    for char in text:
+        page.keyboard.type(char, delay=randint(int(delay * 0.5), int(delay * 1.5)))
+
+
+def human_press_key(page: Page, key: str, count: int = 1, delay: int = 100, add_sleep: bool = True) -> None:
+    """Presses a key with a random delay, optionally sleeping between presses."""
+    for _ in range(count):
+        page.keyboard.press(key, delay=randint(int(delay * 0.5), int(delay * 1.5)))
+        if add_sleep:
+            time.sleep(randint(int(delay * 1.5), int(delay * 2.5)) / 1000)
+
+
+########################################################################################################################
+# CRAWLER CLASS
+
+
+class PlaywrightCrawler:
+    """A robust, proxy-enabled Playwright crawler with captcha bypass and retry logic."""
+
+    def __init__(self, proxy_interface: ProxyInterface):
+        """
+        Initializes the crawler with a proxy interface.
+
+        Args:
+            proxy_interface (ProxyInterface): An object to fetch proxy credentials.
+        """
+        self.proxy_interface = proxy_interface
+        self.pw: Optional[Camoufox] = None
+        self.browser: Optional[Browser] = None
+        self.context: Optional[BrowserContext] = None
+        self.page: Optional[Page] = None
+
+    def __enter__(self) -> Self:
+        """Initializes the browser context when entering the `with` statement."""
+        self.init_context()
+        return self
+
+    def __exit__(
+        self,
+        exc_type: Optional[type[BaseException]],
+        exc_val: Optional[BaseException],
+        exc_tb: Optional[TracebackType],
+    ) -> None:
+        """Safely closes the browser context upon exit."""
+        if self.pw:
+            self.pw.__exit__(exc_type, exc_val, exc_tb)
+
+    @retry(
+        wait=wait_exponential(exp_base=2, multiplier=3, max=90),
+        stop=stop_after_delay(timedelta(minutes=10)),
+        before_sleep=before_sleep_log(logger, logging.INFO),
+        reraise=True,
+    )
+    def init_context(self) -> Self:
+        """Initializes a new browser instance and context with a fresh proxy."""
+        try:
+            host, port, user, pwd = self.proxy_interface.get_proxies(raw=True, use_auth=True)
+            proxy_url = f"http://{host}:{port}"
+            proxy_cfg = {"server": proxy_url}
+
+            if user and pwd:
+                proxy_cfg.update({"username": user, "password": pwd})
+
+            logger.info(f"Starting browser with proxy: {proxy_url}")
+            self.pw = Camoufox(headless=True, geoip=True, humanize=True, proxy=proxy_cfg)
+            self.browser = self.pw.__enter__()
+            self.context = self.browser.new_context()
+            self.page = self.context.new_page()
+        except Exception as e:
+            logger.error(f"Failed to initialize browser context: {e}")
+            if self.pw:
+                self.pw.__exit__(type(e), e, e.__traceback__)
+            raise
+        return self
+
+    def restart_context(self) -> None:
+        """Closes the current browser instance and initializes a new one."""
+        logger.info("Restarting browser context...")
+        if self.pw:
+            self.pw.__exit__(None, None, None)
+        self.init_context()
+
+    @retry(
+        retry=retry_if_exception_type((PlaywrightTimeoutError, PlaywrightError)),
+        wait=wait_exponential(exp_base=2, multiplier=3, max=90),
+        stop=stop_after_delay(timedelta(minutes=10)),
+        before_sleep=before_sleep_log(logger, logging.INFO),
+        before=lambda rs: rs.args[0].restart_context() if rs.attempt_number > 1 else None,
+        reraise=True,
+    )
+    def _goto_with_retry(self, url: str) -> Page:
+        """
+        Navigates to a URL with retries for common Playwright errors.
+        Restarts the browser context on repeated failures.
+        """
+        if not (self.page and not self.page.is_closed()):
+            logger.warning("Page is not available or closed. Restarting context.")
+            self.restart_context()
+
+        # self.page is guaranteed to be valid here by the logic above
+        assert self.page is not None
+        self.page.goto(url, timeout=30000, wait_until="domcontentloaded")
+        return self.page
+
+    def goto(self, url: str) -> Page:
+        """
+        Ensures the browser is initialized and navigates to the given URL.
+        Public wrapper for the internal retry-enabled navigation method.
+        """
+        if not self.page:
+            logger.info("Browser context not found, initializing now...")
+            self.init_context()
+        return self._goto_with_retry(url)

datamarket-0.9.48/src/datamarket/utils/playwright/async_api.py

@@ -1,23 +0,0 @@
-########################################################################################################################
-# IMPORTS
-
-import asyncio
-from random import randint
-
-from playwright.async_api import Page
-
-
-########################################################################################################################
-# FUNCTIONS
-
-
-async def human_type(page: Page, text: str, delay: int = 100):
-    for char in text:
-        await page.keyboard.type(char, delay=randint(int(delay * 0.5), int(delay * 1.5)))  # noqa: S311
-
-
-async def human_press_key(page: Page, key: str, count: int = 1, delay: int = 100, sleep=True):
-    for _ in range(count):
-        await page.keyboard.press(key, delay=randint(int(delay * 0.5), int(delay * 1.5)))  # noqa: S311
-        if sleep:
-            await asyncio.sleep(randint(int(delay * 1.5), int(delay * 2.5)) / 1000)  # noqa: S311

datamarket-0.9.48/src/datamarket/utils/playwright/sync_api.py

@@ -1,23 +0,0 @@
-########################################################################################################################
-# IMPORTS
-
-import time
-from random import randint
-
-from playwright.sync_api import Page
-
-
-########################################################################################################################
-# FUNCTIONS
-
-
-def human_type(page: Page, text: str, delay: int = 100):
-    for char in text:
-        page.keyboard.type(char, delay=randint(int(delay * 0.5), int(delay * 1.5)))  # noqa: S311
-
-
-def human_press_key(page: Page, key: str, count: int = 1, delay: int = 100, sleep=True):
-    for _ in range(count):
-        page.keyboard.press(key, delay=randint(int(delay * 0.5), int(delay * 1.5)))  # noqa: S311
-        if sleep:
-            time.sleep(randint(int(delay * 1.5), int(delay * 2.5)) / 1000)  # noqa: S311