dandiapi.py

"""REST API client for interacting with DANDI Archive instances.

This module provides client classes for communicating with DANDI Archive API
servers, including asset management, dandiset operations, and authentication.

The main classes are:
- DandiAPIClient: High-level client for DANDI API operations
- RESTFullAPIClient: Base HTTP client with retry and authentication
- RemoteDandiset: Represents a dandiset on the server
- RemoteAsset: Represents an asset (file) on the server
"""

from __future__ import annotations

from abc import ABC, abstractmethod
from collections.abc import Callable, Iterable, Iterator, Sequence
from concurrent.futures import ThreadPoolExecutor
from dataclasses import dataclass
from datetime import datetime
from enum import Enum
from fnmatch import fnmatchcase
import json
import os.path
from pathlib import Path, PurePosixPath
import posixpath
import re
from time import sleep, time
from types import TracebackType
from typing import TYPE_CHECKING, Any, Dict, List, Optional

import click
from dandischema import models
import dandischema.consts
from packaging.version import Version as PackagingVersion
from pydantic import BaseModel, Field, PrivateAttr
import requests
import tenacity
from yarl import URL

from . import get_logger
from .consts import (
    DOWNLOAD_TIMEOUT,
    DRAFT,
    MAX_CHUNK_SIZE,
    REQUEST_RETRIES,
    RETRY_STATUSES,
    ZARR_DELETE_BATCH_SIZE,
    DandiInstance,
    EmbargoStatus,
)
from .exceptions import HTTP404Error, NotFoundError, SchemaVersionError
from .keyring_utils import keyring_lookup, keyring_save
from .misctypes import Digest, RemoteReadableAsset
from .utils import (
    USER_AGENT,
    check_dandi_version,
    chunked,
    ensure_datetime,
    get_instance,
    get_retry_after,
    is_interactive,
    is_page2_url,
    joinurl,
)

if TYPE_CHECKING:
    from typing_extensions import Self

    from .apicache import APIMetadataCache


lgr = get_logger()


class AssetType(Enum):
    """
    .. versionadded:: 0.36.0

    An enum for the different kinds of resources that an asset's actual data
    can be
    """

    BLOB = 1
    ZARR = 2


class VersionStatus(Enum):
    PENDING = "Pending"
    VALIDATING = "Validating"
    VALID = "Valid"
    INVALID = "Invalid"
    PUBLISHING = "Publishing"
    PUBLISHED = "Published"


# Following class is loosely based on GirderClient, with authentication etc
# being stripped.
# TODO: add copyright/license info
class RESTFullAPIClient:
    """
    Base class for a JSON-based HTTP(S) client for interacting with a given
    base API URL.

    All request methods can take either an absolute URL or a slash-separated
    path; in the latter case, the path is appended to the base API URL
    (separated by a slash) in order to determine the actual URL to make the
    request of.

    `RESTFullAPIClient` instances are usable as context managers, in which case
    they will close their associated session on exit.
    """

    def __init__(
        self,
        api_url: str,
        session: requests.Session | None = None,
        headers: dict | None = None,
    ) -> None:
        """
        :param str api_url: The base HTTP(S) URL to prepend to request paths
        :param session: an optional `requests.Session` instance to use; if not
            specified, a new session is created
        :param headers: an optional `dict` of headers to send in every request
        """
        self.api_url = api_url
        if session is None:
            session = requests.Session()
        session.headers["User-Agent"] = USER_AGENT
        if headers is not None:
            session.headers.update(headers)
        self.session = session
        #: Default number of items to request per page when paginating (`None`
        #: means to use the server's default)
        self.page_size: int | None = None
        #: How many pages to fetch at once when parallelizing pagination
        self.page_workers: int = 5

    def __enter__(self) -> Self:
        return self

    def __exit__(
        self,
        exc_type: type[BaseException] | None,
        exc_val: BaseException | None,
        exc_tb: TracebackType | None,
    ) -> None:
        self.session.close()

    def request(
        self,
        method: str,
        path: str,
        params: dict | None = None,
        data: Any = None,
        files: dict | None = None,
        json: Any = None,
        headers: dict | None = None,
        json_resp: bool = True,
        retry_statuses: Sequence[int] = (),
        retry_if: Callable[[requests.Response], Any] | None = None,
        **kwargs: Any,
    ) -> Any:
        """
        This method looks up the appropriate method, constructs a request URL
        from the base URL, path, and parameters, and then sends the request. If
        the method is unknown or if the path is not found, an exception is
        raised; otherwise, a JSON object is returned with the response.

        This is a convenience method to use when making basic requests that do
        not involve multipart file data that might need to be specially encoded
        or handled differently.

        :param method: The HTTP method to use in the request (GET, POST, etc.)
        :type method: str
        :param path: A string containing the path elements for this request
        :type path: str
        :param params: A dictionary mapping strings to strings, to be used
            as the key/value pairs in the request parameters.
        :type params: dict
        :param data: A dictionary, bytes or file-like object to send in the
            body.
        :param files: A dictionary of 'name' => file-like-objects for multipart
            encoding upload.
        :type files: dict
        :param json: A JSON object to send in the request body.
        :type json: dict
        :param headers: If present, a dictionary of headers to encode in the
            request.
        :type headers: dict
        :param json_resp: Whether the response should be parsed as JSON. If
            False, the raw response object is returned. To get the raw binary
            content of the response, use the ``content`` attribute of the
            return value, e.g.

            .. code-block:: python

                resp = client.get('my/endpoint', json_resp=False)
                print(resp.content)  # Raw binary content
                print(resp.headers)  # Dict of headers

        :type json_resp: bool
        :param retry_statuses: a sequence of HTTP response status codes to
            retry in addition to `dandi.consts.RETRY_STATUSES`
        :param retry_if: an optional predicate applied to a failed HTTP
            response to test whether to retry
        """

        url = self.get_url(path)

        if headers is None:
            headers = {}
        if json_resp and "accept" not in headers:
            headers["accept"] = "application/json"

        lgr.debug("%s %s", method.upper(), url)

        try:
            for i, attempt in enumerate(
                tenacity.Retrying(
                    wait=tenacity.wait_exponential(exp_base=1.25, multiplier=1.25),
                    # urllib3's ConnectionPool isn't thread-safe, so we
                    # sometimes hit ConnectionErrors on the start of an upload.
                    # Retry when this happens.
                    # Cf. <https://github.com/urllib3/urllib3/issues/951>.
                    retry=tenacity.retry_if_exception_type(
                        (requests.ConnectionError, requests.HTTPError)
                    ),
                    stop=tenacity.stop_after_attempt(REQUEST_RETRIES),
                    reraise=True,
                )
            ):
                with attempt:
                    result = self.session.request(
                        method,
                        url,
                        params=params,
                        data=data,
                        files=files,
                        json=json,
                        headers=headers,
                        **kwargs,
                    )
                    if result.status_code in [*RETRY_STATUSES, *retry_statuses] or (
                        retry_if is not None and retry_if(result)
                    ):
                        if attempt.retry_state.attempt_number < REQUEST_RETRIES:
                            lgr.warning(
                                "Will retry: Error %d while sending %s request to %s: %s",
                                result.status_code,
                                method,
                                url,
                                result.text,
                            )
                            if data is not None and hasattr(data, "seek"):
                                data.seek(0)
                        if retry_after := get_retry_after(result):
                            lgr.debug(
                                "Sleeping for %d seconds as instructed in response "
                                "(in addition to tenacity imposed)",
                                retry_after,
                            )
                            sleep(retry_after)
                        result.raise_for_status()
        except Exception as e:
            if isinstance(e, requests.HTTPError):
                lgr.error(
                    "HTTP request failed repeatedly: Error %d while sending %s request to %s: %s",
                    e.response.status_code if e.response is not None else "?",
                    method,
                    url,
                    e.response.text if e.response is not None else "?",
                )
            else:
                lgr.exception("HTTP connection failed")
            raise

        if i > 0:
            lgr.info(
                "%s %s succeeded after %d retr%s",
                method.upper(),
                url,
                i,
                "y" if i == 1 else "ies",
            )

        lgr.debug("Response: %d", result.status_code)

        # If success, return the json object. Otherwise throw an exception.
        if not result.ok:
            msg = f"Error {result.status_code} while sending {method} request to {url}"
            if result.status_code == 409:
                # Blob exists on server; log at DEBUG level
                lgr.debug("%s: %s", msg, result.text)
            else:
                lgr.error("%s: %s", msg, result.text)
            if len(result.text) <= 1024:
                msg += f": {result.text}"
            else:
                msg += f": {result.text[:1024]}... [{len(result.text)}-char response truncated]"
            if result.status_code == 404:
                raise HTTP404Error(msg, response=result)
            else:
                raise requests.HTTPError(msg, response=result)

        if json_resp:
            if result.text.strip():
                return result.json()
            else:
                return None
        else:
            return result

    def get_url(self, path: str) -> str:
        """
        Append a slash-separated ``path`` to the instance's base URL.  The two
        components are separated by a single slash, removing any excess slashes
        that would be present after naïve concatenation.

        If ``path`` is already an absolute URL, it is returned unchanged.
        """
        return joinurl(self.api_url, path)

    def get(self, path: str, **kwargs: Any) -> Any:
        """
        Convenience method to call `request()` with the 'GET' HTTP method.
        """
        return self.request("GET", path, **kwargs)

    def post(self, path: str, **kwargs: Any) -> Any:
        """
        Convenience method to call `request()` with the 'POST' HTTP method.
        """
        return self.request("POST", path, **kwargs)

    def put(self, path: str, **kwargs: Any) -> Any:
        """
        Convenience method to call `request()` with the 'PUT' HTTP method.
        """
        return self.request("PUT", path, **kwargs)

    def delete(self, path: str, **kwargs: Any) -> Any:
        """
        Convenience method to call `request()` with the 'DELETE' HTTP method.
        """
        return self.request("DELETE", path, **kwargs)

    def patch(self, path: str, **kwargs: Any) -> Any:
        """
        Convenience method to call `request()` with the 'PATCH' HTTP method.
        """
        return self.request("PATCH", path, **kwargs)

    def paginate(
        self,
        path: str,
        page_size: int | None = None,
        params: dict | None = None,
    ) -> Iterator:
        """
        Paginate through the resources at the given path: GET the path, yield
        the values in the ``"results"`` key, and repeat with the URL in the
        ``"next"`` key until it is ``null``.

        If the first ``"next"`` key is the same as the initially-requested URL
        but with the ``page`` query parameter set to ``2``, then the remaining
        pages are fetched concurrently in separate threads, `page_workers`
        (default 5) at a time.  This behavior requires the initial response to
        contain a ``"count"`` key giving the number of items across all pages.

        :param page_size:
            If non-`None`, overrides the client's `page_size` attribute for
            this sequence of pages
        """
        if page_size is None:
            page_size = self.page_size
        if page_size is not None:
            if params is None:
                params = {}
            params["page_size"] = page_size

        resp = self.get(path, params=params, json_resp=False)
        r = resp.json()
        if r["next"] is not None:
            page1 = resp.history[0].url if resp.history else resp.url
            if not is_page2_url(page1, r["next"]):
                if os.environ.get("DANDI_PAGINATION_DISABLE_FALLBACK"):
                    raise RuntimeError(
                        f"API server changed pagination strategy: {page1} URL"
                        f" is now followed by {r['next']}"
                    )
                else:
                    while True:
                        yield from r["results"]
                        if r.get("next"):
                            r = self.get(r["next"])
                        else:
                            return
        yield from r["results"]
        if r["next"] is None:
            return

        if page_size is None:
            page_size = len(r["results"])
        pages = (r["count"] + page_size - 1) // page_size

        def get_page(pageno: int) -> list:
            params2 = params.copy() if params is not None else {}
            params2["page"] = pageno
            results = self.get(path, params=params2)["results"]
            assert isinstance(results, list)
            return results

        with ThreadPoolExecutor(max_workers=self.page_workers) as pool:
            futures = [pool.submit(get_page, i) for i in range(2, pages + 1)]
            try:
                for f in futures:
                    yield from f.result()
            finally:
                for f in futures:
                    f.cancel()


class DandiAPIClient(RESTFullAPIClient):
    """A client for interacting with a DANDI API server"""

    def __init__(
        self,
        api_url: str | None = None,
        token: str | None = None,
        dandi_instance: DandiInstance | None = None,
        cache: bool = False,
    ) -> None:
        """
        Construct a client instance for the given API URL or DANDI instance
        (mutually exclusive options).  If no URL or instance is supplied, the
        instance specified by the :envvar:`DANDI_INSTANCE` environment variable
        (default value: ``"dandi"``) is used.

        :param str api_url: Base API URL of the server to interact with.
            - For DANDI production, use  ``"https://api.dandiarchive.org/api"``
            - For DANDI sandbox, use
              ``"https://api.sandbox.dandiarchive.org/api"``
        :param str token: User API Key. Note that different instance APIs have
            different keys.
        :param bool cache: When ``True``, API metadata responses are cached
            persistently to disk (in an sqlite3 database) and validated against
            ``modified`` timestamps.  Controlled by the :envvar:`DANDI_CACHE`
            environment variable (``"ignore"`` disables, ``"clear"`` wipes the
            cache on first access).
        """
        check_dandi_version()
        if api_url is None:
            if dandi_instance is None:
                instance_name = os.environ.get("DANDI_INSTANCE", "dandi")
                dandi_instance = get_instance(instance_name)
            api_url = dandi_instance.api
        elif dandi_instance is not None:
            raise ValueError(
                "api_url and dandi_instance are mutually exclusive. "
                "Use either 'api_url' to specify a custom API URL, "
                "or 'dandi_instance' to use a registered DANDI instance, but not both."
            )
        else:
            dandi_instance = get_instance(api_url)
        super().__init__(api_url)
        self.dandi_instance: DandiInstance = dandi_instance
        if token is not None:
            self.authenticate(token)
        if cache:
            from .apicache import APIMetadataCache

            self._cache: APIMetadataCache | None = APIMetadataCache()
        else:
            self._cache = None

    @property
    def cache(self) -> APIMetadataCache | None:
        """The persistent API metadata cache, or ``None`` if caching is disabled."""
        return self._cache

    @classmethod
    def for_dandi_instance(
        cls,
        instance: str | DandiInstance,
        token: str | None = None,
        authenticate: bool = False,
    ) -> DandiAPIClient:
        """
        Construct a client instance for the server identified by ``instance``
        (either the name of a registered DANDI instance or a
        `DandiInstance` instance) and an optional authentication token/API key.
        If no token is supplied and ``authenticate`` is true,
        `dandi_authenticate()` is called on the instance before returning it.
        """
        client = cls(dandi_instance=get_instance(instance), token=token)
        if token is None and authenticate:
            client.dandi_authenticate()
        return client

    def authenticate(self, token: str, save_to_keyring: bool = False) -> None:
        """
        Set the authentication token/API key used by the `DandiAPIClient`.
        Before setting the token, a test request to ``/auth/token`` is made to
        check the token's validity; if it fails, a `requests.HTTPError` is
        raised.

        If ``save_to_keyring`` is true, then (after querying ``/auth/token``
        but before setting the API key used by the client), the token is saved
        in the user's keyring at the same location as used by
        `dandi_authenticate()`.

        .. versionchanged:: 0.53.0

            ``save_to_keyring`` added
        """
        # Fails if token is invalid:
        self.get("/auth/token", headers={"Authorization": f"token {token}"})
        if save_to_keyring:
            keyring_save(self._get_keyring_ids()[1], "key", token)
            lgr.debug("Stored key in keyring")
        self.session.headers["Authorization"] = f"token {token}"

    def dandi_authenticate(self) -> None:
        """
        Acquire and set the authentication token/API key used by the
        `DandiAPIClient`.
        If the :envvar:`{INSTANCE_NAME}_API_KEY` environment variable is set, its value
        is used as the token. Here, ``{INSTANCE_NAME}`` is the uppercased instance name
        with hyphens replaced by underscores. Otherwise, the token is looked up in the
        user's keyring under the service ":samp:`dandi-api-{self.dandi_instance.name}`"
        [#auth]_ and username "``key``". If no token is found there, the user is
        prompted for the token, and, if it proves to be valid, it is stored in the
        user's keyring.

        .. [#auth] E.g., "``dandi-api-dandi``" for the production server or
                   "``dandi-api-dandi-sandbox``" for the sandbox server
        """
        # Shortcut for advanced folks
        env_var_name = self.api_key_env_var
        api_key = os.environ.get(env_var_name, None)
        if api_key:
            lgr.debug(f"Using `{env_var_name}` environment variable as the API key")
            self.authenticate(api_key)
            return
        client_name, app_id = self._get_keyring_ids()
        keyring_backend, api_key = keyring_lookup(app_id, "key")
        key_from_keyring = api_key is not None
        while True:
            if not api_key:
                api_key = input(f"Please provide API Key for {client_name}: ")
                key_from_keyring = False
            try:
                lgr.debug(
                    "Using API key from %s",
                    {True: "keyring", False: "user input"}[key_from_keyring],
                )
                self.authenticate(api_key)
            except requests.HTTPError:
                if is_interactive() and click.confirm(
                    "API key is invalid; enter another?"
                ):
                    api_key = None
                    continue
                else:
                    raise
            else:
                if not key_from_keyring:
                    keyring_backend.set_password(app_id, "key", api_key)
                    lgr.debug("Stored key in keyring")
                break

    def _get_keyring_ids(self) -> tuple[str, str]:
        client_name = self.dandi_instance.name
        return (client_name, f"dandi-api-{client_name}")

    @property
    def _instance_id(self) -> str:
        return self.dandi_instance.name.upper()

    def get_dandiset(
        self, dandiset_id: str, version_id: str | None = None, lazy: bool = True
    ) -> RemoteDandiset:
        """
        Fetches the Dandiset with the given ``dandiset_id``.  If ``version_id``
        is not specified, the `RemoteDandiset`'s version is set to the most
        recent published version if there is one, otherwise to the draft
        version.

        If ``lazy`` is true, no requests are actually made until any data is
        requested from the `RemoteDandiset`.
        """
        if lazy:
            return RemoteDandiset(self, dandiset_id, version_id)
        else:
            try:
                d = RemoteDandiset.from_data(
                    self, self.get(f"/dandisets/{dandiset_id}/")
                )
            except HTTP404Error:
                raise NotFoundError(
                    f"No such Dandiset: {dandiset_id!r}. "
                    "Verify the Dandiset ID is correct and that you have access. "
                )
            if version_id is not None and version_id != d.version_id:
                if version_id == DRAFT:
                    return d.for_version(d.draft_version)
                else:
                    return d.for_version(version_id)
            return d

    def get_dandisets(
        self,
        *,
        draft: bool | None = None,
        embargoed: bool | None = None,
        empty: bool | None = None,
        mine: bool | None = None,
        order: str | None = None,
        search: str | None = None,
    ) -> Iterator[RemoteDandiset]:
        """
        Returns a generator of all Dandisets on the server.  For each Dandiset,
        the `RemoteDandiset`'s version is set to the most recent published
        version if there is one, otherwise to the draft version.

        .. versionchanged:: 0.61.0

            ``draft``, ``embargoed``, ``empty``, ``mine``, ``order``, and
            ``search`` parameters added

        :param draft:
            If true, Dandisets that have only draft versions (i.e., that
            haven't yet been published) will be included in the results
            (default true)

        :param embargoed:
            If true, embargoed Dandisets will be included in the results
            (default false)

        :param empty:
            If true, empty Dandisets will be included in the results (default
            true)

        :param mine:
            If true, only Dandisets owned by the authenticated user will be
            retrieved (default false)

        :param order:
            The field to sort the results by.  The accepted field names are
            ``"id"``, ``"name"``, ``"modified"``, and ``"size"``.  Prepend a
            hyphen to the field name to reverse the sort order.

        :param search:
            A search string to filter the returned Dandisets by.  The string is
            searched for in the metadata of Dandiset versions.
        """
        for data in self.paginate(
            "/dandisets/",
            params={
                "draft": draft,
                "embargoed": embargoed,
                "empty": empty,
                "ordering": order,
                "search": search,
                "user": "me" if mine else None,
            },
        ):
            yield RemoteDandiset.from_data(self, data)

    def create_dandiset(
        self, name: str, metadata: dict[str, Any], *, embargo: bool = False
    ) -> RemoteDandiset:
        """
        Creates a Dandiset with the given name & metadata.  If ``embargo`` is
        `True`, the resulting Dandiset will be embargoed.

        .. versionchanged:: 0.61.0

            ``embargo`` argument added
        """
        return RemoteDandiset.from_data(
            self,
            self.post(
                "/dandisets/",
                json={"name": name, "metadata": metadata},
                params={"embargo": "true" if embargo else "false"},
            ),
        )

    def check_schema_version(self, schema_version: str | None = None) -> None:
        """
        Confirms that the given schema version at the client is "compatible" with the server.

        Compatibility here means that the server's schema version can be either

        - lower than client has, but within the same MAJOR.MINOR component of the version
          number for 0.x series, and same MAJOR version for/after 1.x series;
        - the same;
        - higher than the client has, but only if the client's schema version is listed
          among the server's `allowed_schema_versions` (as returned by the `/info` API endpoint),
          or if not there -- `dandischema.consts.ALLOWED_INPUT_SCHEMAS` is consulted.

        If neither of above, a `SchemaVersionError` is raised.

        :param schema_version: the schema version to be confirmed for compatibility with the server;
           if not set, the schema version for the installed ``dandischema`` library is used.
        """
        if schema_version is None:
            schema_version = models.get_schema_version()
        server_info = self.get("/info/")
        server_schema_version = server_info.get("schema_version")
        if not server_schema_version:
            raise RuntimeError(
                "Server did not provide schema_version in /info/;"
                f" returned {server_info!r}"
            )
        server_ver, our_ver = PackagingVersion(server_schema_version), PackagingVersion(
            schema_version
        )
        if server_ver > our_ver:
            # TODO: potentially adjust here if name would be different: see
            # https://github.com/dandi/dandi-archive/issues/2624
            allowed_schema_versions = server_info.get(
                "allowed_schema_versions", dandischema.consts.ALLOWED_INPUT_SCHEMAS
            )
            if schema_version not in allowed_schema_versions:
                raise SchemaVersionError(
                    f"Server uses schema version {server_schema_version};"
                    f" client only supports prior {schema_version} and it"
                    f" is not among any of the allowed upgradable schema versions"
                    f" ({', '.join(allowed_schema_versions)}) .  You may need to"
                    " upgrade dandi and/or dandischema."
                )

            # TODO: check current server behavior which is likely to just not care!
            # So that is where server might need to provide support for upgrades upon
            # providing metadata.
        elif (
            server_ver.major == 0 and server_ver.release[:2] != our_ver.release[:2]
        ) or (
            server_ver.major != our_ver.major
        ):  # MAJOR, MINOR within 0.x.y and MAJOR within 1.x.y
            raise SchemaVersionError(
                f"Server uses older incompatible schema version {server_schema_version};"
                f" client supports {schema_version}."
            )
        elif server_ver < our_ver:
            # Compatible older server version -- all good, but inform the user
            # TODO: potentially downgrade the record to match the schema,
            #       see https://github.com/dandi/dandi-schema/issues/343
            lgr.warning(
                "Server uses schema version %s older than client's %s (dandischema library %s). "
                "Server might fail to validate such assets and you might not be able to "
                "publish this dandiset until server is upgraded. "
                "Alternatively, you may downgrade dandischema and reupload.",
                server_ver,
                our_ver,
                dandischema.__version__,
            )

    def get_asset(self, asset_id: str) -> BaseRemoteAsset:
        """
        Fetch the asset with the given asset ID.  If the given asset does not
        exist, a `NotFoundError` is raised.

        The returned object will not have any information about the Dandiset
        associated with the asset; for that, the `RemoteDandiset.get_asset()`
        method must be used instead.
        """
        try:
            info = self.get(f"/assets/{asset_id}/info/")
        except HTTP404Error:
            raise NotFoundError(
                f"No such asset: {asset_id!r}. "
                "Verify the asset ID is correct. "
                "Use 'dandi ls' to list available assets."
            )
        metadata = info.pop("metadata", None)
        return BaseRemoteAsset.from_base_data(self, info, metadata)

    @property
    def api_key_env_var(self) -> str:
        """
        Get the name of the environment variable that can be used to specify the
        API key for the associated DANDI instance.
        """
        return f"{self.dandi_instance.name.upper().replace('-', '_')}_API_KEY"


# `arbitrary_types_allowed` is needed for `client: DandiAPIClient`
class APIBase(BaseModel, populate_by_name=True, arbitrary_types_allowed=True):
    """
    Base class for API objects implemented in pydantic.

    This class (aside from the `json_dict()` method) is an implementation
    detail; do not rely on it.
    """

    def json_dict(self) -> dict[str, Any]:
        """
        Convert to a JSONable `dict`, omitting the ``client`` attribute and
        using the same field names as in the API
        """
        return self.model_dump(mode="json", by_alias=True)


class Version(APIBase):
    """
    The version information for a Dandiset retrieved from the API.

    Stringifying a `Version` returns its identifier.

    This class should not be instantiated by end-users directly.  Instead,
    instances should be retrieved from the appropriate attributes & methods of
    `RemoteDandiset`.
    """

    #: The version identifier
    identifier: str = Field(alias="version")
    #: The name of the version
    name: str
    #: The number of assets in the version
    asset_count: int
    #: The total size in bytes of all assets in the version
    size: int
    #: The timestamp at which the version was created
    created: datetime
    #: The timestamp at which the version was last modified
    modified: datetime
    status: VersionStatus

    def __str__(self) -> str:
        return self.identifier


class RemoteValidationError(APIBase):
    """
    .. versionadded:: 0.49.0

    Validation error record obtained from a server.  Not to be confused with
    :class:`dandi.validate_types.ValidationResult`, which provides richer
    representation of validation errors.
    """

    field: str
    message: str


class RemoteAssetValidationError(RemoteValidationError):
    path: Optional[str] = None


class VersionInfo(Version):
    """
    .. versionadded:: 0.49.0

    Version information for a Dandiset, including information about validation
    errors
    """

    asset_validation_errors: List[RemoteAssetValidationError]
    version_validation_errors: List[RemoteValidationError]


class RemoteDandisetData(APIBase):
    """
    Class for storing the data for a Dandiset retrieved from the API.

    This class is an implementation detail and should not be used by third
    parties.
    """

    identifier: str
    created: datetime
    modified: datetime
    contact_person: str
    embargo_status: EmbargoStatus
    most_recent_published_version: Optional[Version] = None
    draft_version: Version


class RemoteDandiset:
    """
    Representation of a Dandiset (as of a certain version) retrieved from the
    API.

    Stringifying a `RemoteDandiset` returns a string of the form
    :samp:`"{server_id}:{dandiset_id}/{version_id}"`.

    This class should not be instantiated by end-users directly.  Instead,
    instances should be retrieved from the appropriate attributes & methods of
    `DandiAPIClient` and `RemoteDandiset`.
    """

    def __init__(
        self,
        client: DandiAPIClient,
        identifier: str,
        version: str | Version | None = None,
        data: dict[str, Any] | RemoteDandisetData | None = None,
    ) -> None:
        #: The `DandiAPIClient` instance that returned this `RemoteDandiset`
        #: and which the latter will use for API requests
        self.client: DandiAPIClient = client
        #: The Dandiset identifier
        self.identifier: str = identifier
        self._version_id: str | None
        self._version: Version | None
        if version is None:
            self._version_id = None
            self._version = None
        elif isinstance(version, str):
            self._version_id = version
            self._version = None
        else:
            self._version_id = version.identifier
            self._version = version
        self._data: RemoteDandisetData | None
        if data is not None:
            self._data = RemoteDandisetData.model_validate(data)
        else:
            self._data = None

    def __str__(self) -> str:
        return f"{self.client._instance_id}:{self.identifier}/{self.version_id}"

    def _get_data(self) -> RemoteDandisetData:
        if self._data is None:
            try:
                self._data = RemoteDandisetData.model_validate(
                    self.client.get(f"/dandisets/{self.identifier}/")
                )
            except HTTP404Error:
                raise NotFoundError(f"No such Dandiset: {self.identifier}")
        return self._data

    @property
    def version_id(self) -> str:
        """The identifier for the Dandiset version"""
        if self._version_id is None:
            self._version_id = self.version.identifier
        return self._version_id

    @property
    def version(self) -> Version:
        """The version in question of the Dandiset"""
        if self._version is None:
            if self._version_id is None:
                self._get_data()
            if self._data is not None:
                for v in [
                    self._data.most_recent_published_version,
                    self._data.draft_version,
                ]:
                    if v is not None and (
                        self._version_id is None or v.identifier == self.version_id
                    ):
                        self._version = v
                        self._version_id = v.identifier
                        return v
            assert self._version_id is not None
            self._version = self.get_version(self._version_id)
        return self._version

    @property
    def created(self) -> datetime:
        """The timestamp at which the Dandiset was created"""
        return self._get_data().created

    @property
    def modified(self) -> datetime:
        """The timestamp at which the Dandiset was last modified"""
        return self._get_data().modified

    @property
    def contact_person(self) -> str:
        """The name of the registered contact person for the Dandiset"""
        return self._get_data().contact_person

    @property
    def embargo_status(self) -> EmbargoStatus:
        """The current embargo status for the Dandiset"""
        return self._get_data().embargo_status

    @property
    def most_recent_published_version(self) -> Version | None:
        """
        The most recent published (non-draft) version of the Dandiset, or
        `None` if no versions have been published
        """
        return self._get_data().most_recent_published_version

    @property
    def draft_version(self) -> Version:
        """The draft version of the Dandiset"""
        return self._get_data().draft_version

    @property
    def api_path(self) -> str:
        """
        The path (relative to the base endpoint for the DANDI API) at
        which API requests for interacting with the Dandiset itself are made
        """
        return f"/dandisets/{self.identifier}/"

    @property
    def api_url(self) -> str:
        """
        The URL at which API requests for interacting with the Dandiset itself
        are made
        """
        return self.client.get_url(self.api_path)

    @property
    def version_api_path(self) -> str:
        """
        The path (relative to the base endpoint for the DANDI API) at
        which API requests for interacting with the version in question of the
        Dandiset are made
        """
        return f"/dandisets/{self.identifier}/versions/{self.version_id}/"

    @property
    def version_api_url(self) -> str:
        """
        The URL at which API requests for interacting with the version in
        question of the Dandiset are made
        """
        return self.client.get_url(self.version_api_path)

    @classmethod
    def from_data(cls, client: DandiAPIClient, data: dict[str, Any]) -> RemoteDandiset:
        """
        Construct a `RemoteDandiset` instance from a `DandiAPIClient` and a
        `dict` of raw string fields in the same format as returned by the API.
        If the ``"most_recent_published_version"`` field is set, that is used
        as the Dandiset's version; otherwise, ``"draft_version"`` is used.

        This is a low-level method that non-developers would normally only use
        when acquiring data using means outside of this library.
        """
        if data.get("most_recent_published_version") is not None:
            version = Version.model_validate(data["most_recent_published_version"])
        else:
            version = Version.model_validate(data["draft_version"])
        return cls(
            client=client, identifier=data["identifier"], version=version, data=data
        )

    @staticmethod
    def _normalize_path(path: str) -> str:
        """
        Helper to normalize path before passing it to the server.

        We and API call it "path" but it is really a "prefix" with inherent
        semantics of containing directory divider '/' and referring to a
        directory when terminated with '/'.
        """
        # Server (now) expects path to be a proper prefix, so to account for user
        # possibly specifying ./ or some other relative paths etc, let's normalize
        # the path.
        # Ref: https://github.com/dandi/dandi-cli/issues/1452
        path_normed = posixpath.normpath(path)
        if path_normed == ".":
            path_normed = ""
        elif path.endswith("/"):
            # we need to make sure that we have a trailing slash if we had it before
            path_normed += "/"
        if path_normed != path:
            lgr.debug("Normalized path %r to %r", path, path_normed)
        return path_normed

    def json_dict(self) -> dict[str, Any]:
        """
        Convert to a JSONable `dict`, omitting the ``client`` attribute and
        using the same field names as in the API
        """
        return {
            **self._get_data().json_dict(),
            "version": self.version.json_dict(),
        }

    def refresh(self) -> None:
        """
        Update the `RemoteDandiset` in-place with the latest data from the
        server.  The `RemoteDandiset` continues to have the same version as
        before, but the cached version data is internally cleared and may be
        different upon subsequent access.
        """
        self._data = None
        self._get_data()
        # Clear _version so it will be refetched the next time it is accessed
        self._version = None

    def get_versions(self, order: str | None = None) -> Iterator[Version]:
        """
        Returns an iterator of all available `Version`\\s for the Dandiset

        Versions can be sorted by a given field by passing the name of that
        field as the ``order`` parameter.  Currently, the only accepted field
        name is ``"created"``.  Prepend a hyphen to the field name to reverse
        the sort order.
        """
        try:
            for v in self.client.paginate(
                f"{self.api_path}versions/", params={"order": order}
            ):
                yield Version.model_validate(v)
        except HTTP404Error:
            raise NotFoundError(f"No such Dandiset: {self.identifier!r}")

    def get_version(self, version_id: str) -> VersionInfo:
        """
        Get information about a given version of the Dandiset.  If the given
        version does not exist, a `NotFoundError` is raised.

        .. versionchanged:: 0.49.0

            This method now returns a `VersionInfo` instance instead of just a
            `Version`.
        """
        try:
            return VersionInfo.model_validate(
                self.client.get(
                    f"/dandisets/{self.identifier}/versions/{version_id}/info/"
                )
            )
        except HTTP404Error:
            raise NotFoundError(
                f"No such version: {version_id!r} of Dandiset {self.identifier}"
            )

    def for_version(self, version_id: str | Version) -> RemoteDandiset:
        """
        Returns a copy of the `RemoteDandiset` with the `version` attribute set
        to the given `Version` object or the `Version` with the given version
        ID.  If a version ID given and the version does not exist, a
        `NotFoundError` is raised.
        """
        if isinstance(version_id, str):
            version_id = self.get_version(version_id)
        return type(self)(
            client=self.client,
            identifier=self.identifier,
            version=version_id,
            data=self._data,
        )

    def delete(self) -> None:
        """
        Delete the Dandiset from the server.  Any further access of the
        instance's data attributes afterwards will result in a `NotFoundError`.
        """
        self.client.delete(self.api_path)
        self._data = None
        self._version = None

    def get_metadata(self) -> models.Dandiset:
        """
        Fetch the metadata for this version of the Dandiset as a
        `dandischema.models.Dandiset` instance

        .. note::

            Only published Dandiset versions can be expected to have valid
            metadata.  Consider using `get_raw_metadata()` instead in order to
            fetch unstructured, possibly-invalid metadata.
        """
        return models.Dandiset.model_validate(self.get_raw_metadata())

    def get_raw_metadata(self) -> dict[str, Any]:
        """
        Fetch the metadata for this version of the Dandiset as an unprocessed
        `dict`.

        When the client was created with ``cache=True``, results are served
        from a persistent on-disk cache whenever the version's ``modified``
        timestamp has not changed.
        """
        cache = self.client.cache
        entity_id = f"{self.identifier}/{self.version_id}"
        if cache is not None and self._version is not None:
            modified = self._version.modified.isoformat()
            cached = cache.get(self.client.api_url, "dandiset", entity_id, modified)
            if cached is not None:
                return cached
        try:
            data = self.client.get(self.version_api_path)
            assert isinstance(data, dict)
        except HTTP404Error:
            raise NotFoundError(f"No such asset: {self}")
        if cache is not None:
            modified = self.version.modified.isoformat()
            cache.set(self.client.api_url, "dandiset", entity_id, modified, data)
        return data

    def set_metadata(self, metadata: models.Dandiset) -> None:
        """
        Set the metadata for this version of the Dandiset to the given value
        """
        self.set_raw_metadata(metadata.model_dump(mode="json", exclude_none=True))

    def set_raw_metadata(self, metadata: dict[str, Any]) -> None:
        """
        Set the metadata for this version of the Dandiset to the given value
        """
        self.client.put(
            self.version_api_path,
            json={"metadata": metadata, "name": metadata.get("name", "")},
        )

    def wait_until_valid(self, max_time: float = 120) -> None:
        """
        Wait at most ``max_time`` seconds for the Dandiset to be valid for
        publication.  If the Dandiset does not become valid in time, a
        `ValueError` is raised.
        """
        lgr.debug("Waiting for Dandiset %s to complete validation ...", self.identifier)
        start = time()
        while time() - start < max_time:
            v = self.get_version(self.version_id)
            if v.status is VersionStatus.VALID and not v.asset_validation_errors:
                return
            sleep(0.5)
        # TODO: Improve the presentation of the error messages
        about = {
            "asset_validation_errors": [
                e.json_dict() for e in v.asset_validation_errors
            ],
            "version_validation_errors": [
                e.json_dict() for e in v.version_validation_errors
            ],
        }
        raise ValueError(
            f"Dandiset {self.identifier} is {v.status.value}: {json.dumps(about, indent=4)}"
        )

    def publish(self, max_time: float = 120) -> RemoteDandiset:
        """
        Publish the draft version of the Dandiset and wait at most ``max_time``
        seconds for the publication operation to complete.  If the operation
        does not complete in time, a `ValueError` is raised.

        Returns a copy of the `RemoteDandiset` with the `version` attribute set
        to the new published `Version`.
        """
        draft_api_path = f"/dandisets/{self.identifier}/versions/draft/"
        self.client.post(f"{draft_api_path}publish/")
        lgr.debug(
            "Waiting for Dandiset %s to complete publication ...", self.identifier
        )
        start = time()
        while time() - start < max_time:
            v = Version.model_validate(self.client.get(f"{draft_api_path}info/"))
            if v.status is VersionStatus.PUBLISHED:
                break
            sleep(0.5)
        else:
            raise ValueError(f"Dandiset {self.identifier} did not publish in time")
        for v in self.get_versions(order="-created"):
            return self.for_version(v)
        raise AssertionError(
            f"No published versions found for Dandiset {self.identifier}"
        )

    def get_assets(self, order: str | None = None) -> Iterator[RemoteAsset]:
        """
        Returns an iterator of all assets in this version of the Dandiset.

        Assets can be sorted by a given field by passing the name of that field
        as the ``order`` parameter.  The accepted field names are
        ``"created"``, ``"modified"``, and ``"path"``.  Prepend a hyphen to the
        field name to reverse the sort order.
        """
        try:
            for a in self.client.paginate(
                f"{self.version_api_path}assets/", params={"order": order}
            ):
                yield RemoteAsset.from_data(self, a)
        except HTTP404Error:
            raise NotFoundError(
                f"No such version: {self.version_id!r} of Dandiset {self.identifier}"
            )

    def get_asset(self, asset_id: str) -> RemoteAsset:
        """
        Fetch the asset in this version of the Dandiset with the given asset
        ID.  If the given asset does not exist, a `NotFoundError` is raised.
        """
        try:
            info = self.client.get(f"{self.version_api_path}assets/{asset_id}/info/")
        except HTTP404Error:
            raise NotFoundError(f"No such asset: {asset_id!r} for {self}")
        metadata = info.pop("metadata", None)
        return RemoteAsset.from_data(self, info, metadata)

    def get_assets_with_path_prefix(
        self, path: str, order: str | None = None
    ) -> Iterator[RemoteAsset]:
        """
        Returns an iterator of all assets in this version of the Dandiset whose
        `~RemoteAsset.path` attributes start with ``path``

        ``path`` is normalized first to possibly remove leading ``./`` or relative
        paths (e.g., ``../``) within it.

        Assets can be sorted by a given field by passing the name of that field
        as the ``order`` parameter.  The accepted field names are
        ``"created"``, ``"modified"``, and ``"path"``.  Prepend a hyphen to the
        field name to reverse the sort order.
        """
        try:
            for a in self.client.paginate(
                f"{self.version_api_path}assets/",
                params={"path": self._normalize_path(path), "order": order},
            ):
                yield RemoteAsset.from_data(self, a)
        except HTTP404Error:
            raise NotFoundError(
                f"No such version: {self.version_id!r} of Dandiset {self.identifier}"
            )

    def get_assets_by_glob(
        self, pattern: str, order: str | None = None
    ) -> Iterator[RemoteAsset]:
        """
        .. versionadded:: 0.44.0

        Returns an iterator of all assets in this version of the Dandiset whose
        `~RemoteAsset.path` attributes match the glob pattern ``pattern``

        Assets can be sorted by a given field by passing the name of that field
        as the ``order`` parameter.  The accepted field names are
        ``"created"``, ``"modified"``, and ``"path"``.  Prepend a hyphen to the
        field name to reverse the sort order.
        """
        try:
            for a in self.client.paginate(
                f"{self.version_api_path}assets/",
                params={"glob": pattern, "order": order},
            ):
                yield RemoteAsset.from_data(self, a)
        except HTTP404Error:
            raise NotFoundError(
                f"No such version: {self.version_id!r} of Dandiset {self.identifier}"
            )

    def get_asset_by_path(self, path: str) -> RemoteAsset:
        """
        Fetch the asset in this version of the Dandiset whose
        `~RemoteAsset.path` equals ``path``.  If the given asset does not
        exist, a `NotFoundError` is raised.

        ``path`` is normalized first to possibly remove leading ``./`` or relative
        paths (e.g., ``../``) within it.
        """
        path = self._normalize_path(path)
        try:
            # Weed out any assets that happen to have the given path as a
            # proper prefix:
            (asset,) = (
                a for a in self.get_assets_with_path_prefix(path) if a.path == path
            )
        except ValueError:
            raise NotFoundError(
                f"No asset at path {path!r} in version {self.version_id}. "
                "Verify the path is correct and the asset exists in this version. "
                "Use 'dandi ls' to list available assets."
            )
        else:
            return asset

    def download_directory(
        self,
        assets_dirpath: str,
        dirpath: str | Path,
        chunk_size: int = MAX_CHUNK_SIZE,
    ) -> None:
        """
        Download all assets under the virtual directory ``assets_dirpath`` to
        the directory ``dirpath``.  Downloads are synchronous.

        ``path`` is normalized first to possibly remove leading ``./`` or relative
        paths (e.g., ``../``) within it.
        """
        if assets_dirpath and not assets_dirpath.endswith("/"):
            assets_dirpath += "/"
        # need to normalize explicitly since we do use it below also
        # to deduce portion of the path to strip off.
        assets_dirpath = self._normalize_path(assets_dirpath)
        assets = list(self.get_assets_with_path_prefix(assets_dirpath))
        for a in assets:
            filepath = Path(dirpath, a.path[len(assets_dirpath) :])
            filepath.parent.mkdir(parents=True, exist_ok=True)
            a.download(filepath, chunk_size=chunk_size)

    def upload_raw_asset(
        self,
        filepath: str | Path,
        asset_metadata: dict[str, Any],
        jobs: int | None = None,
        replace_asset: RemoteAsset | None = None,
    ) -> RemoteAsset:
        """
        Upload the file at ``filepath`` with metadata ``asset_metadata`` to
        this version of the Dandiset and return the resulting asset.  Blocks
        until the upload is complete.

        .. deprecated:: 0.36.0
            Use the ``upload()`` method of `~dandi.files.LocalAsset` instances
            instead

        :param filepath: the path to the local file to upload
        :type filepath: str or PathLike
        :param dict asset_metadata:
            Metadata for the uploaded asset file.  Must include a "path" field
            giving the forward-slash-separated path at which the uploaded file
            will be placed on the server.
        :param int jobs: Number of threads to use for uploading; defaults to 5
        :param RemoteAsset replace_asset: If set, replace the given asset,
            which must have the same path as the new asset
        """
        # Avoid circular import by importing within function:
        from .files import LocalAsset, dandi_file

        df = dandi_file(filepath)
        if not isinstance(df, LocalAsset):
            raise ValueError(f"{filepath}: not an asset file")
        return df.upload(
            self, metadata=asset_metadata, jobs=jobs, replacing=replace_asset
        )

    def iter_upload_raw_asset(
        self,
        filepath: str | Path,
        asset_metadata: dict[str, Any],
        jobs: int | None = None,
        replace_asset: RemoteAsset | None = None,
    ) -> Iterator[dict]:
        """
        Upload the file at ``filepath`` with metadata ``asset_metadata`` to
        this version of the Dandiset, returning a generator of status
        `dict`\\s.

        .. deprecated:: 0.36.0
            Use the ``iter_upload()`` method of `~dandi.files.LocalAsset`
            instances instead

        :param filepath: the path to the local file to upload
        :type filepath: str or PathLike
        :param dict asset_metadata:
            Metadata for the uploaded asset file.  Must include a "path" field
            giving the forward-slash-separated path at which the uploaded file
            will be placed on the server.
        :param int jobs:
            Number of threads to use for uploading; defaults to 5
        :param RemoteAsset replace_asset: If set, replace the given asset,
            which must have the same path as the new asset
        :returns:
            A generator of `dict`\\s containing at least a ``"status"`` key.
            Upon successful upload, the last `dict` will have a status of
            ``"done"`` and an ``"asset"`` key containing the resulting
            `RemoteAsset`.
        """
        # Avoid circular import by importing within function:
        from .files import LocalAsset, dandi_file

        df = dandi_file(filepath)
        if not isinstance(df, LocalAsset):
            raise ValueError(f"{filepath}: not an asset file")
        return df.iter_upload(
            self, metadata=asset_metadata, jobs=jobs, replacing=replace_asset
        )


class BaseRemoteAsset(ABC, APIBase):
    """
    Representation of an asset retrieved from the API without associated
    Dandiset information.

    This is an abstract class; its concrete subclasses are
    `BaseRemoteBlobAsset` (for assets backed by blobs) and
    `BaseRemoteZarrAsset` (for assets backed by Zarrs).

    Stringifying a `BaseRemoteAsset` returns a string of the form
    :samp:`"{server_id}:asset/{asset_id}"`.

    This class should not be instantiated by end-users directly.  Instead,
    instances should be retrieved from the appropriate methods of
    `DandiAPIClient`.
    """

    #: The `DandiAPIClient` instance that returned this `BaseRemoteAsset`
    #: and which the latter will use for API requests
    client: DandiAPIClient = Field(exclude=True)
    #: The asset identifier
    identifier: str = Field(alias="asset_id")
    #: The asset's (forward-slash-separated) path
    path: str
    #: The size of the asset in bytes
    size: int
    #: The date at which the asset was created
    created: datetime
    #: The date at which the asset was last modified
    modified: datetime
    #: Metadata supplied at initialization; returned when metadata is requested
    #: instead of performing an API call
    _metadata: Optional[Dict[str, Any]] = PrivateAttr(default=None)

    def __init__(self, **data: Any) -> None:  # type: ignore[no-redef]
        super().__init__(**data)
        # Pydantic insists on not initializing any attributes that start with
        # underscores, so we have to do it ourselves.
        self._metadata = data.get("metadata", data.get("_metadata"))

    def __eq__(self, other: Any) -> bool:
        if type(self) is type(other):
            # dict() includes fields with `exclude=True` (which are absent from
            # the return value of `model_dump()`) but not private fields.  We
            # want to compare the former but not the latter.
            return dict(self) == dict(other)
        else:
            return NotImplemented

    def __str__(self) -> str:
        return f"{self.client._instance_id}:assets/{self.identifier}"

    @classmethod
    def from_base_data(
        cls,
        client: DandiAPIClient,
        data: dict[str, Any],
        metadata: dict[str, Any] | None = None,
    ) -> BaseRemoteAsset:
        """
        Construct a `BaseRemoteAsset` instance from a `DandiAPIClient`, a
        `dict` of raw data in the same format as returned by the API's
        pagination endpoints, and optional raw asset metadata.

        This is a low-level method that non-developers would normally only use
        when acquiring data using means outside of this library.
        """
        klass: type[BaseRemoteAsset]
        if data.get("blob") is not None:
            klass = BaseRemoteBlobAsset
            if data.pop("zarr", None) is not None:
                raise ValueError("Asset data contains both `blob` and `zarr`'")
        elif data.get("zarr") is not None:
            klass = BaseRemoteZarrAsset
            if data.pop("blob", None) is not None:
                raise ValueError("Asset data contains both `blob` and `zarr`'")
        else:
            raise ValueError("Asset data contains neither `blob` nor `zarr`")
        return klass(client=client, **data, _metadata=metadata)  # type: ignore[call-arg]

    @property
    def api_path(self) -> str:
        """
        The path (relative to the base endpoint for the DANDI API) at
        which API requests for interacting with the asset itself are made
        """
        return f"/assets/{self.identifier}/"

    @property
    def api_url(self) -> str:
        """
        The URL at which API requests for interacting with the asset itself are
        made
        """
        return self.client.get_url(self.api_path)

    @property
    def base_download_url(self) -> str:
        """
        The URL from which the asset can be downloaded, sans any Dandiset
        identifiers (cf. `RemoteAsset.download_url`)
        """
        return self.client.get_url(f"/assets/{self.identifier}/download/")

    def get_metadata(self) -> models.Asset:
        """
        Fetch the metadata for the asset as a `dandischema.models.Asset`
        instance

        .. note::

            Only assets in published Dandiset versions can be expected to have
            valid metadata.  Consider using `get_raw_metadata()` instead in
            order to fetch unstructured, possibly-invalid metadata.
        """
        return models.Asset.model_validate(self.get_raw_metadata())

    def get_raw_metadata(self) -> dict[str, Any]:
        """Fetch the metadata for the asset as an unprocessed `dict`.

        When the client was created with ``cache=True``, results are served
        from a persistent on-disk cache whenever the asset's ``modified``
        timestamp has not changed.
        """
        if self._metadata is not None:
            return self._metadata
        cache = self.client.cache
        modified = self.modified.isoformat()
        if cache is not None:
            cached = cache.get(self.client.api_url, "asset", self.identifier, modified)
            if cached is not None:
                return cached
        try:
            data = self.client.get(self.api_path)
            assert isinstance(data, dict)
        except HTTP404Error:
            raise NotFoundError(f"No such asset: {self}")
        if cache is not None:
            cache.set(self.client.api_url, "asset", self.identifier, modified, data)
        return data

    def get_raw_digest(self, digest_type: str | models.DigestType | None = None) -> str:
        """
        Retrieves the value of the given type of digest from the asset's
        metadata.  Raises `NotFoundError` if there is no entry for the given
        digest type.

        If no digest type is specified, the same type as used by `get_digest()`
        is returned.

        .. versionchanged:: 0.36.0
            Renamed from ``get_digest()`` to ``get_raw_digest()``
        """
        if digest_type is None:
            digest_type = self.digest_type.value
        elif isinstance(digest_type, models.DigestType):
            digest_type = digest_type.value
        metadata = self.get_raw_metadata()
        try:
            digest = metadata["digest"][digest_type]
        except KeyError:
            raise NotFoundError(f"No {digest_type} digest found in metadata")
        assert isinstance(digest, str)
        return digest

    def get_digest(self) -> Digest:
        """
        .. versionadded:: 0.36.0
            Replaces the previous version of ``get_digest()``, now renamed to
            `get_raw_digest()`

        Retrieves the DANDI etag digest of the appropriate type for the asset:
        a dandi-etag digest for blob resources or a dandi-zarr-checksum for
        Zarr resources
        """
        algorithm = self.digest_type
        return Digest(algorithm=algorithm, value=self.get_raw_digest(algorithm))

    def get_content_url(
        self,
        regex: str = r".*",
        follow_redirects: bool | int = False,
        strip_query: bool = False,
    ) -> str:
        """
        Returns a URL for downloading the asset, found by inspecting the
        metadata; specifically, returns the first ``contentUrl`` that matches
        ``regex``.  Raises `NotFoundError` if the metadata does not contain a
        matching URL.

        If ``follow_redirects`` is `True`, a ``HEAD`` request is made to
        resolve any & all redirects before returning the URL.  If
        ``follow_redirects`` is an integer, at most that many redirects are
        followed.

        If ``strip_query`` is true, any query parameters are removed from the
        final URL before returning it.
        """
        url: str
        for url in self.get_raw_metadata().get("contentUrl", []):
            if re.search(regex, url):
                break
        else:
            raise NotFoundError(
                "No matching URL found in asset's contentUrl metadata field"
            )
        try:
            if follow_redirects is True:
                url = self.client.request(
                    "HEAD", url, json_resp=False, allow_redirects=True
                ).url
            elif follow_redirects:
                for _ in range(follow_redirects):
                    r = self.client.request(
                        "HEAD", url, json_resp=False, allow_redirects=False
                    )
                    if "Location" in r.headers:
                        url = r.headers["Location"]
                    else:
                        break
        except requests.HTTPError as e:
            if e.request is not None and isinstance(e.request.url, str):
                url = e.request.url
            else:
                raise  # reraise since we need to figure out how to handle such a case
        if strip_query:
            url = str(URL(url).with_query(None))
        return url

    def get_download_file_iter(
        self, chunk_size: int = MAX_CHUNK_SIZE
    ) -> Callable[[int], Iterator[bytes]]:
        """
        Returns a function that when called (optionally with an offset into the
        asset to start downloading at) returns a generator of chunks of the
        asset.

        :raises ValueError: if the asset is not backed by a blob
        """
        if self.asset_type is not AssetType.BLOB:
            raise ValueError(
                f"Cannot download asset {self} directly: asset is of type"
                f" {self.asset_type.name}, not BLOB"
            )

        url = self.base_download_url

        def downloader(start_at: int = 0) -> Iterator[bytes]:
            lgr.debug("Starting download from %s", url)
            headers = None
            if start_at > 0:
                headers = {"Range": f"bytes={start_at}-"}
            result = self.client.session.get(
                url, stream=True, headers=headers, timeout=DOWNLOAD_TIMEOUT
            )
            # TODO: apparently we might need retries here as well etc
            # if result.status_code not in (200, 201):
            result.raise_for_status()
            nbytes, nchunks = 0, 0
            for chunk in result.iter_content(chunk_size=chunk_size):
                nchunks += 1
                if chunk:  # could be some "keep alive"?
                    nbytes += len(chunk)
                    yield chunk
                else:
                    lgr.debug("'Empty' chunk downloaded for %s", url)
            lgr.info(
                "Asset %s (%d bytes in %d chunks starting from %d) successfully "
                "downloaded from %s",
                self.identifier,
                nbytes,
                nchunks,
                start_at,
                url,
            )

        return downloader

    def download(self, filepath: str | Path, chunk_size: int = MAX_CHUNK_SIZE) -> None:
        """
        Download the asset to ``filepath``.  Blocks until the download is
        complete.

        :raises ValueError: if the asset is not backed by a blob
        """
        downloader = self.get_download_file_iter(chunk_size=chunk_size)
        with open(filepath, "wb") as fp:
            for chunk in downloader(0):
                fp.write(chunk)

    @property
    @abstractmethod
    def asset_type(self) -> AssetType:
        """
        .. versionadded:: 0.36.0

        The type of the asset's underlying data
        """
        ...

    @property
    def digest_type(self) -> models.DigestType:
        """
        .. versionadded:: 0.36.0

        The primary digest algorithm used by DANDI for the asset,
        determined based on its underlying data: dandi-etag for blob resources,
        dandi-zarr-checksum for Zarr resources
        """
        if self.asset_type is AssetType.ZARR:
            return models.DigestType.dandi_zarr_checksum
        else:
            return models.DigestType.dandi_etag


class BaseRemoteBlobAsset(BaseRemoteAsset):
    """
    .. versionadded:: 0.36.0

    A `BaseRemoteAsset` whose actual data is a blob resource
    """

    #: The ID of the underlying blob resource
    blob: str

    @property
    def asset_type(self) -> AssetType:
        """
        .. versionadded:: 0.36.0

        The type of the asset's underlying data
        """
        return AssetType.BLOB

    def as_readable(self) -> RemoteReadableAsset:
        """
        .. versionadded:: 0.50.0

        Returns a `Readable` instance that can be used to obtain a file-like
        object for reading bytes directly from the asset on the server
        """
        md = self.get_raw_metadata()
        local_prefix = self.client.api_url.lower()
        for url in md.get("contentUrl", []):
            if not url.lower().startswith(local_prefix):
                # This must be the S3 URL
                try:
                    size = int(md["contentSize"])
                except (KeyError, TypeError, ValueError):
                    lgr.warning('"contentSize" not set for asset %s', self.identifier)
                    r = requests.head(url)
                    r.raise_for_status()
                    size = int(r.headers["Content-Length"])
                mtime: datetime | None
                try:
                    mtime = ensure_datetime(md["blobDateModified"])
                except (KeyError, TypeError, ValueError):
                    mtime = None
                name = PurePosixPath(md["path"]).name
                return RemoteReadableAsset(url=url, size=size, mtime=mtime, name=name)
        raise NotFoundError("S3 URL not found in asset's contentUrl metadata field")


class BaseRemoteZarrAsset(BaseRemoteAsset):
    """
    .. versionadded:: 0.36.0

    A `BaseRemoteAsset` whose actual data is a Zarr resource
    """

    #: The ID of the underlying Zarr resource
    zarr: str

    @property
    def asset_type(self) -> AssetType:
        """
        .. versionadded:: 0.36.0

        The type of the asset's underlying data
        """
        return AssetType.ZARR

    def iterfiles(self, prefix: str | None = None) -> Iterator[RemoteZarrEntry]:
        """
        Returns a generator of all `RemoteZarrEntry`\\s within the Zarr,
        optionally limited to those whose path starts with the given prefix
        """
        for r in self.client.paginate(
            f"{self.client.api_url}/zarr/{self.zarr}/files", params={"prefix": prefix}
        ):
            data = ZarrEntryServerData.model_validate(r)
            yield RemoteZarrEntry.from_server_data(self, data)

    def get_entry_by_path(self, path: str) -> RemoteZarrEntry:
        """
        Fetch the entry in this Zarr whose `~RemoteZarrEntry.path` equals
        ``path``.  If the given entry does not exist, a `NotFoundError` is
        raised.
        """
        try:
            # Weed out any entries that happen to have the given path as a
            # proper prefix:
            (entry,) = (e for e in self.iterfiles(prefix=path) if str(e) == path)
        except ValueError:
            raise NotFoundError(f"No entry at path {path!r}")
        else:
            return entry

    def rmfiles(self, files: Iterable[RemoteZarrEntry], reingest: bool = True) -> None:
        """
        Delete one or more files from the Zarr.

        If ``reingest`` is true, after performing the deletion, the client
        triggers a recalculation of the Zarr's checksum and waits for it to
        complete.
        """
        # Don't bother checking that the entries are actually files or even
        # belong to this Zarr, as if they're not, the server will return an
        # error anyway.
        for entries in chunked(files, ZARR_DELETE_BATCH_SIZE):
            self.client.delete(
                f"/zarr/{self.zarr}/files/",
                json=[{"path": str(e)} for e in entries],
            )
        if reingest:
            self.client.post(f"/zarr/{self.zarr}/finalize/")
            while True:
                sleep(2)
                r = self.client.get(f"/zarr/{self.zarr}/")
                if r["status"] == "Complete":
                    break


class RemoteAsset(BaseRemoteAsset):
    """
    Subclass of `BaseRemoteAsset` that includes information about the Dandiset
    to which the asset belongs.

    This is an abstract class; its concrete subclasses are `RemoteBlobAsset`
    (for assets backed by blobs) and `RemoteZarrAsset` (for assets backed by
    Zarrs).

    This class should not be instantiated by end-users directly.  Instead,
    instances should be retrieved from the appropriate attributes & methods of
    `RemoteDandiset`.
    """

    #: The identifier for the Dandiset to which the asset belongs
    dandiset_id: str = Field(exclude=True)

    #: The identifier for the version of the Dandiset to which the asset
    #: belongs
    version_id: str = Field(exclude=True)

    @classmethod
    def from_data(
        cls,
        dandiset: RemoteDandiset,
        data: dict[str, Any],
        metadata: dict[str, Any] | None = None,
    ) -> RemoteAsset:
        """
        Construct a `RemoteAsset` instance from a `RemoteDandiset`, a `dict` of
        raw data in the same format as returned by the API's pagination
        endpoints, and optional raw asset metadata.

        This is a low-level method that non-developers would normally only use
        when acquiring data using means outside of this library.
        """
        klass: type[RemoteAsset]
        if data.get("blob") is not None:
            klass = RemoteBlobAsset
            if data.pop("zarr", None) is not None:
                raise ValueError("Asset data contains both `blob` and `zarr`'")
        elif data.get("zarr") is not None:
            klass = RemoteZarrAsset
            if data.pop("blob", None) is not None:
                raise ValueError("Asset data contains both `blob` and `zarr`'")
        else:
            raise ValueError("Asset data contains neither `blob` nor `zarr`")
        return klass(  # type: ignore[call-arg]
            client=dandiset.client,
            dandiset_id=dandiset.identifier,
            version_id=dandiset.version_id,
            **data,
            _metadata=metadata,
        )

    @property
    def api_path(self) -> str:
        """
        The path (relative to the base endpoint for the DANDI API) at
        which API requests for interacting with the asset itself are made
        """
        return f"/dandisets/{self.dandiset_id}/versions/{self.version_id}/assets/{self.identifier}/"

    @property
    def api_url(self) -> str:
        """
        The URL at which API requests for interacting with the asset itself are
        made
        """
        return self.client.get_url(self.api_path)

    @property
    def download_url(self) -> str:
        """
        The URL from which the asset can be downloaded, including Dandiset
        identifiers (cf. `BaseRemoteAsset.base_download_url`)
        """
        return self.client.get_url(f"{self.api_path}download/")

    def set_metadata(self, metadata: models.Asset) -> None:
        """
        Set the metadata for the asset to the given value and update the
        `RemoteAsset` in place.
        """
        return self.set_raw_metadata(
            metadata.model_dump(mode="json", exclude_none=True)
        )

    @abstractmethod
    def set_raw_metadata(self, metadata: dict[str, Any]) -> None:
        """
        Set the metadata for the asset on the server to the given value and
        update the `RemoteAsset` in place.
        """
        ...

    def rename(self, dest: str) -> None:
        """
        .. versionadded:: 0.41.0

        Change the path of the asset on the server to the given value and
        update the `RemoteAsset` in place.  If another asset already exists at
        the given path, a `requests.HTTPError` is raised.
        """
        md = self.get_raw_metadata().copy()
        md["path"] = dest
        self.set_raw_metadata(md)

    def delete(self) -> None:
        """Delete the asset"""
        self.client.delete(self.api_path)


class RemoteBlobAsset(RemoteAsset, BaseRemoteBlobAsset):
    """
    .. versionadded:: 0.36.0

    A `RemoteAsset` whose actual data is a blob resource
    """

    def set_raw_metadata(self, metadata: dict[str, Any]) -> None:
        """
        Set the metadata for the asset on the server to the given value and
        update the `RemoteBlobAsset` in place.
        """
        data = self.client.put(
            self.api_path, json={"metadata": metadata, "blob_id": self.blob}
        )
        self.identifier = data["asset_id"]
        self.path = data["path"]
        self.size = int(data["size"])
        self.created = ensure_datetime(data["created"])
        self.modified = ensure_datetime(data["modified"])
        self._metadata = data["metadata"]


class RemoteZarrAsset(RemoteAsset, BaseRemoteZarrAsset):
    """
    .. versionadded:: 0.36.0

    A `RemoteAsset` whose actual data is a Zarr resource
    """

    def set_raw_metadata(self, metadata: dict[str, Any]) -> None:
        """
        Set the metadata for the asset on the server to the given value and
        update the `RemoteZarrAsset` in place.
        """
        data = self.client.put(
            self.api_path, json={"metadata": metadata, "zarr_id": self.zarr}
        )
        self.identifier = data["asset_id"]
        self.path = data["path"]
        self.size = int(data["size"])
        self.created = ensure_datetime(data["created"])
        self.modified = ensure_datetime(data["modified"])
        self._metadata = data["metadata"]


@dataclass
class RemoteZarrEntry:
    """
    .. versionadded:: 0.36.0

    A file within a `RemoteZarrAsset`

    .. versionchanged:: 0.48.0

        - No longer represents directories
        - No longer implements `~dandi.misctypes.BasePath`
    """

    #: The `DandiAPIClient` instance used for API requests
    client: DandiAPIClient
    #: The ID of the Zarr backing the asset
    zarr_id: str
    #: The path components of the entry
    parts: tuple[str, ...]
    #: The timestamp at which the entry was last modified
    modified: datetime
    #: The entry's digest
    digest: Digest
    #: The entry's size in bytes
    size: int

    @classmethod
    def from_server_data(
        cls, asset: BaseRemoteZarrAsset, data: ZarrEntryServerData
    ) -> RemoteZarrEntry:
        """:meta private:"""
        return cls(
            client=asset.client,
            zarr_id=asset.zarr,
            parts=tuple(data.key.split("/")),
            modified=data.last_modified,
            digest=Digest(algorithm=models.DigestType.md5, value=data.etag),
            size=data.size,
        )

    def __str__(self) -> str:
        return "/".join(self.parts)

    @property
    def name(self) -> str:
        """The basename of the path object"""
        assert self.parts
        return self.parts[-1]

    @property
    def suffix(self) -> str:
        """The final file extension of the basename, if any"""
        i = self.name.rfind(".")
        if 0 < i < len(self.name) - 1:
            return self.name[i:]
        else:
            return ""

    @property
    def suffixes(self) -> list[str]:
        """A list of the basename's file extensions"""
        if self.name.endswith("."):
            return []
        name = self.name.lstrip(".")
        return ["." + suffix for suffix in name.split(".")[1:]]

    @property
    def stem(self) -> str:
        """The basename without its final file extension, if any"""
        i = self.name.rfind(".")
        if 0 < i < len(self.name) - 1:
            return self.name[:i]
        else:
            return self.name

    def match(self, pattern: str) -> bool:
        """Tests whether the path matches the given glob pattern"""
        if pattern.startswith("/"):
            raise ValueError(f"Absolute paths not allowed: {pattern!r}")
        patparts = tuple(q for q in pattern.split("/") if q)
        if not patparts:
            raise ValueError("Empty pattern")
        if len(patparts) > len(self.parts):
            return False
        for part, pat in zip(reversed(self.parts), reversed(patparts)):
            if not fnmatchcase(part, pat):
                return False
        return True

    @property
    def download_url(self) -> str:
        """
        .. versionadded:: 0.67.0

        The URL from which the entry can be downloaded
        """
        return str(
            URL(self.client.get_url(f"/zarr/{self.zarr_id}/files/")).with_query(
                {"prefix": str(self), "download": "true"}
            )
        )

    def get_download_file_iter(
        self, chunk_size: int = MAX_CHUNK_SIZE
    ) -> Callable[[int], Iterator[bytes]]:
        """
        Returns a function that when called (optionally with an offset into the
        file to start downloading at) returns a generator of chunks of the file
        """
        url = self.download_url

        def downloader(start_at: int = 0) -> Iterator[bytes]:
            lgr.debug("Starting download from %s", url)
            headers = None
            if start_at > 0:
                headers = {"Range": f"bytes={start_at}-"}
            result = self.client.session.get(
                url, stream=True, headers=headers, timeout=DOWNLOAD_TIMEOUT
            )
            # TODO: apparently we might need retries here as well etc
            # if result.status_code not in (200, 201):
            result.raise_for_status()
            for chunk in result.iter_content(chunk_size=chunk_size):
                if chunk:  # could be some "keep alive"?
                    yield chunk
            lgr.info("File %s in Zarr %s successfully downloaded", self, self.zarr_id)

        return downloader


class ZarrEntryServerData(BaseModel):
    """
    Intermediate structure used for parsing details on a Zarr entry returned by
    the server

    :meta private:
    """

    key: str = Field(alias="Key")
    last_modified: datetime = Field(alias="LastModified")
    etag: str = Field(alias="ETag")
    size: int = Field(alias="Size")