Use REST APIs to resolve DOIs + cleanup dataverse provider

.github/workflows/test.yml

@@ -60,6 +60,7 @@ jobs:
           - r
           - unit
           - venv
+          - contentproviders
         include:
           # The actions/setup-python action with Python version 3.6 isn't
           # possible to use with the ubuntu-22.04 runner, so we use ubuntu-20.04

repo2docker/contentproviders/dataverse.py

@@ -1,9 +1,11 @@
+import hashlib
 import json
 import os
 import shutil
-from urllib.parse import parse_qs, urlparse, urlunparse
+from typing import List
+from urllib.parse import parse_qs, urlparse
 
-from ..utils import copytree, deep_get
+from ..utils import copytree, deep_get, is_doi
 from .doi import DoiProvider
 
 
@@ -23,10 +25,11 @@ def __init__(self):
             self.hosts = json.load(fp)["installations"]
         super().__init__()
 
-    def detect(self, doi, ref=None, extra_args=None):
-        """Trigger this provider for things that resolve to a Dataverse dataset.
+    def detect(self, spec, ref=None, extra_args=None):
+        """
+        Detect if given spec is hosted on dataverse
 
-        Handles:
+        The spec can be:
         - DOI pointing to {siteURL}/dataset.xhtml?persistentId={persistentId}
         - DOI pointing to {siteURL}/file.xhtml?persistentId={persistentId}&...
         - URL {siteURL}/api/access/datafile/{fileId}
@@ -35,9 +38,11 @@ def detect(self, doi, ref=None, extra_args=None):
         - https://dataverse.harvard.edu/api/access/datafile/3323458
         - doi:10.7910/DVN/6ZXAGT
         - doi:10.7910/DVN/6ZXAGT/3YRRYJ
-
         """
-        url = self.doi2url(doi)
+        if is_doi(spec):
+            url = self.doi2url(spec)
+        else:
+            url = spec
         # Parse the url, to get the base for later API calls
         parsed_url = urlparse(url)
 
@@ -53,54 +58,117 @@ def detect(self, doi, ref=None, extra_args=None):
         if host is None:
             return
 
-        query_args = parse_qs(parsed_url.query)
-        # Corner case handling
-        if parsed_url.path.startswith("/file.xhtml"):
-            # There's no way of getting file information using its persistentId, the only thing we can do is assume that doi
-            # is structured as "doi:<dataset_doi>/<file_doi>" and try to handle dataset that way.
-            new_doi = doi.rsplit("/", 1)[0]
-            if new_doi == doi:
-                # tough luck :( Avoid inifite recursion and exit.
-                return
-            return self.detect(new_doi)
-        elif parsed_url.path.startswith("/api/access/datafile"):
-            # Raw url pointing to a datafile is a typical output from an External Tool integration
-            entity_id = os.path.basename(parsed_url.path)
-            search_query = "q=entityId:" + entity_id + "&type=file"
-            # Knowing the file identifier query search api to get parent dataset
-            search_url = urlunparse(
-                parsed_url._replace(path="/api/search", query=search_query)
+        # Used only for content_id
+        self.url = url
+
+        # At this point, we *know* this is a dataverse URL, because:
+        # 1. The DOI resolved to a particular host (if using DOI)
+        # 2. The host is in the list of known dataverse installations
+        #
+        # We don't know exactly what kind of dataverse object this is, but
+        # that can be figured out during fetch as needed
+        return {"host": host, "url": url}
+
+    def get_dataset_id_from_file_id(self, host: str, file_id: str) -> str:
+        """
+        Return the persistent_id (DOI) that a given file_id (int or doi) belongs to
+        """
+        if file_id.isdigit():
+            # the file_id is an integer, rather than a persistent id (DOI)
+            api_url = f"{host}/api/files/{file_id}?returnDatasetVersion=true"
+        else:
+            # the file_id is a doi itself
+            api_url = f"{host}/api/files/:persistentId?persistentId={file_id}&returnDatasetVersion=true"
+
+        resp = self._request(api_url)
+        if resp.status_code == 404:
+            raise ValueError(f"File with id {file_id} not found in {host}")
+
+        resp.raise_for_status()
+
+        data = resp.json()["data"]
+        return data["datasetVersion"]["datasetPersistentId"]
+
+    def get_datafiles(self, dataverse_host: str, url: str) -> List[dict]:
+        """
+        Return a list of dataFiles for given persistent_id
+
+        Supports the following *dataset* URL styles:
+        - /citation: https://dataverse.harvard.edu/citation?persistentId=doi:10.7910/DVN/TJCLKP
+        - /dataset.xhtml: https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/TJCLKP
+
+        Supports the following *file* URL styles:
+        - /api/access/datafile: https://dataverse.harvard.edu/api/access/datafile/3323458
+
+        Supports a subset of the following *file* URL styles:
+        - /file.xhtml: https://dataverse.harvard.edu/file.xhtml?persistentId=doi:10.7910/DVN/6ZXAGT/3YRRYJ
+
+        If a URL can not be parsed, throw an exception
+        """
+
+        parsed_url = urlparse(url)
+        path = parsed_url.path
+        qs = parse_qs(parsed_url.query)
+        dataverse_host = f"{parsed_url.scheme}://{parsed_url.netloc}"
+        url_kind = None
+        persistent_id = None
+        is_ambiguous = False
+
+        # https://dataverse.harvard.edu/citation?persistentId=doi:10.7910/DVN/TJCLKP
+        if path.startswith("/citation"):
+            is_ambiguous = True
+            persistent_id = qs["persistentId"][0]
+        # https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/TJCLKP
+        elif path.startswith("/dataset.xhtml"):
+            #  https://dataverse.harvard.edu/api/access/datafile/3323458
+            persistent_id = qs["persistentId"][0]
+        elif path.startswith("/api/access/datafile"):
+            # What we have here is an entity id, which we can use to get a persistentId
+            file_id = os.path.basename(parsed_url.path)
+            persistent_id = self.get_dataset_id_from_file_id(dataverse_host, file_id)
+        elif parsed_url.path.startswith("/file.xhtml"):
+            file_persistent_id = qs["persistentId"][0]
+            persistent_id = self.get_dataset_id_from_file_id(
+                dataverse_host, file_persistent_id
+            )
+        else:
+            raise ValueError(
+                f"Could not determine persistent id for dataverse URL {url}"
+            )
+
+        dataset_api_url = (
+            f"{dataverse_host}/api/datasets/:persistentId?persistentId={persistent_id}"
+        )
+        resp = self._request(dataset_api_url, headers={"accept": "application/json"})
+        if resp.status_code == 404 and is_ambiguous:
+            # It's possible this is a *file* persistent_id, not a dataset one
+            persistent_id = self.get_dataset_id_from_file_id(
+                dataverse_host, persistent_id
+            )
+            dataset_api_url = f"{dataverse_host}/api/datasets/:persistentId?persistentId={persistent_id}"
+            resp = self._request(
+                dataset_api_url, headers={"accept": "application/json"}
             )
-            self.log.debug("Querying Dataverse: " + search_url)
-            data = self.urlopen(search_url).json()["data"]
-            if data["count_in_response"] != 1:
-                self.log.debug(
-                    f"Dataverse search query failed!\n - doi: {doi}\n - url: {url}\n - resp: {json.dump(data)}\n"
-                )
-                return
-
-            self.record_id = deep_get(data, "items.0.dataset_persistent_id")
-        elif (
-            parsed_url.path.startswith("/dataset.xhtml")
-            and "persistentId" in query_args
-        ):
-            self.record_id = deep_get(query_args, "persistentId.0")
-
-        if hasattr(self, "record_id"):
-            return {"record": self.record_id, "host": host}
+
+            if resp.status_code == 404:
+                # This persistent id is just not here
+                raise ValueError(f"{persistent_id} on {dataverse_host} is not found")
+
+        # We already handled 404, raise error for everything else
+        resp.raise_for_status()
+
+        data = resp.json()["data"]
+
+        return data["latestVersion"]["files"]
 
     def fetch(self, spec, output_dir, yield_output=False):
         """Fetch and unpack a Dataverse dataset."""
-        record_id = spec["record"]
+        url = spec["url"]
         host = spec["host"]
 
-        yield f"Fetching Dataverse record {record_id}.\n"
-        url = f'{host["url"]}/api/datasets/:persistentId?persistentId={record_id}'
-
-        resp = self.urlopen(url, headers={"accept": "application/json"})
-        record = resp.json()["data"]
+        yield f"Fetching Dataverse record {url}.\n"
 
-        for fobj in deep_get(record, "latestVersion.files"):
+        for fobj in self.get_datafiles(host["url"], url):
             file_url = (
                 # without format=original you get the preservation format (plain text, tab separated)
                 f'{host["url"]}/api/access/datafile/{deep_get(fobj, "dataFile.id")}?format=original'
@@ -129,4 +197,4 @@ def fetch(self, spec, output_dir, yield_output=False):
     @property
     def content_id(self):
         """The Dataverse persistent identifier."""
-        return self.record_id
+        return hashlib.sha256(self.url.encode()).hexdigest()

repo2docker/contentproviders/doi.py

@@ -46,21 +46,28 @@ def doi2url(self, doi):
         # Transform a DOI to a URL
         # If not a doi, assume we have a URL and return
         if is_doi(doi):
-            doi = normalize_doi(doi)
-
-            try:
-                resp = self._request(f"https://doi.org/{doi}")
-                resp.raise_for_status()
-            except HTTPError as e:
-                # If the DOI doesn't exist, just return URL
-                if e.response.status_code == 404:
-                    return doi
-                # Reraise any other errors because if the DOI service is down (or
-                # we hit a rate limit) we don't want to silently continue to the
-                # default Git provider as this leads to a misleading error.
-                self.log.error(f"DOI {doi} does not resolve: {e}")
+            normalized_doi = normalize_doi(doi)
+
+            # Use the doi.org resolver API
+            # documented at https://www.doi.org/the-identifier/resources/factsheets/doi-resolution-documentation#5-proxy-server-rest-api
+            req_url = f"https://doi.org/api/handles/{normalized_doi}"
+            resp = self._request(req_url)
+            if resp.status_code == 404:
+                # Not a doi, return what we were passed in
+                return doi
+            elif resp.status_code == 200:
+                data = resp.json()
+                # Pick the first URL we find from the doi response
+                for v in data["values"]:
+                    if v["type"] == "URL":
+                        return v["data"]["value"]
+
+                # No URLs found for this doi, what do we do?
+                self.log.error("DOI {normalized_doi} doesn't point to any URLs")
+                return doi
+            else:
+                # If we get any other status codes, raise error
                 raise
-            return resp.url
         else:
             # Just return what is actulally just a URL
             return doi