xml_upload.py

"""
This module handles the import of XML data into the DSP platform.
"""
import base64
import json
import os
import re
import uuid
from collections import namedtuple
from datetime import datetime
from pathlib import Path
from typing import Optional, cast, Tuple, Any
from urllib.parse import quote_plus

import pandas as pd
from lxml import etree

from knora.dsplib.models.connection import Connection
from knora.dsplib.models.helpers import BaseError
from knora.dsplib.models.permission import Permissions
from knora.dsplib.models.projectContext import ProjectContext
from knora.dsplib.models.resource import ResourceInstanceFactory, ResourceInstance, KnoraStandoffXmlEncoder
from knora.dsplib.models.sipi import Sipi
from knora.dsplib.models.value import KnoraStandoffXml
from knora.dsplib.models.xmlpermission import XmlPermission
from knora.dsplib.models.xmlproperty import XMLProperty
from knora.dsplib.models.xmlresource import XMLResource
from knora.dsplib.utils.shared import try_network_action, validate_xml_against_schema

MetricRecord = namedtuple("MetricRecord", ["res_id", "filetype", "filesize_mb", "event", "duration_ms", "mb_per_sec"])


def _remove_circular_references(resources: list[XMLResource], verbose: bool) -> \
        tuple[list[XMLResource],
              dict[XMLResource, dict[XMLProperty, dict[str, KnoraStandoffXml]]],
              dict[XMLResource, dict[XMLProperty, list[str]]]
              ]:
    """
    Temporarily removes problematic resource-references from a list of resources. A reference is problematic if
    it creates a circle (circular references).

    Args:
        resources: list of resources that possibly contain circular references
        verbose: verbose output if True

    Returns:
        list: list of cleaned resources
        stashed_xml_texts: dict with the stashed XML texts
        stashed_resptr_props: dict with the stashed resptr-props
    """

    if verbose:
        print("Checking resources for unresolvable references...")

    stashed_xml_texts: dict[XMLResource, dict[XMLProperty, dict[str, KnoraStandoffXml]]] = {}
    stashed_resptr_props: dict[XMLResource, dict[XMLProperty, list[str]]] = {}

    # sort the resources according to outgoing resptrs
    ok_resources: list[XMLResource] = []
    nok_resources: list[XMLResource] = []
    ok_res_ids: list[str] = []
    cnt = 0
    nok_len = 9999999
    while len(resources) > 0 and cnt < 10000:
        for resource in resources:
            resptrs = resource.get_resptrs()
            if len(resptrs) == 0:
                ok_resources.append(resource)
                ok_res_ids.append(resource.id)
            else:
                ok = True
                for resptr in resptrs:
                    if resptr not in ok_res_ids:
                        ok = False
                if ok:
                    ok_resources.append(resource)
                    ok_res_ids.append(resource.id)
                else:
                    nok_resources.append(resource)
        resources = nok_resources
        if len(nok_resources) == nok_len:
            # there are circular references. go through all problematic resources, and stash the problematic references.
            nok_resources, ok_res_ids, ok_resources, stashed_xml_texts, stashed_resptr_props = _stash_circular_references(
                nok_resources,
                ok_res_ids,
                ok_resources,
                stashed_xml_texts,
                stashed_resptr_props
            )
        nok_len = len(nok_resources)
        nok_resources = []
        cnt += 1
        if verbose:
            print(f'{cnt}. ordering pass finished.')
    return ok_resources, stashed_xml_texts, stashed_resptr_props


def _stash_circular_references(
    nok_resources: list[XMLResource],
    ok_res_ids: list[str],
    ok_resources: list[XMLResource],
    stashed_xml_texts: dict[XMLResource, dict[XMLProperty, dict[str, KnoraStandoffXml]]],
    stashed_resptr_props: dict[XMLResource, dict[XMLProperty, list[str]]]
) -> Tuple[
    list[XMLResource],
    list[str],
    list[XMLResource],
    dict[XMLResource, dict[XMLProperty, dict[str, KnoraStandoffXml]]],
    dict[XMLResource, dict[XMLProperty, list[str]]]
]:
    for res in nok_resources.copy():
        for link_prop in res.get_props_with_links():
            if link_prop.valtype == 'text':
                for value in link_prop.values:
                    if value.resrefs and not all([_id in ok_res_ids for _id in value.resrefs]):
                        # stash this XML text, replace it by its hash, and remove the
                        # problematic resrefs from the XMLValue's resrefs list
                        value_hash = str(hash(f'{value.value}{datetime.now()}'))
                        if res not in stashed_xml_texts:
                            stashed_xml_texts[res] = {link_prop: {value_hash: cast(KnoraStandoffXml, value.value)}}
                        elif link_prop not in stashed_xml_texts[res]:
                            stashed_xml_texts[res][link_prop] = {value_hash: cast(KnoraStandoffXml, value.value)}
                        else:
                            stashed_xml_texts[res][link_prop][value_hash] = cast(KnoraStandoffXml, value.value)
                        value.value = KnoraStandoffXml(value_hash)
                        value.resrefs = [_id for _id in value.resrefs if _id in ok_res_ids]
            elif link_prop.valtype == 'resptr':
                for value in link_prop.values.copy():
                    if value.value not in ok_res_ids:
                        # value.value is the id of the target resource. stash it, then delete it
                        if res not in stashed_resptr_props:
                            stashed_resptr_props[res] = {}
                            stashed_resptr_props[res][link_prop] = [str(value.value)]
                        else:
                            if link_prop not in stashed_resptr_props[res]:
                                stashed_resptr_props[res][link_prop] = [str(value.value)]
                            else:
                                stashed_resptr_props[res][link_prop].append(str(value.value))
                        link_prop.values.remove(value)
            else:
                raise BaseError(f'ERROR in remove_circular_references(): link_prop.valtype is '
                                f'neither text nor resptr.')

            if len(link_prop.values) == 0:
                # if all values of a link property have been stashed, the property needs to be removed
                res.properties.remove(link_prop)

        ok_resources.append(res)
        ok_res_ids.append(res.id)
        nok_resources.remove(res)

    return nok_resources, ok_res_ids, ok_resources, stashed_xml_texts, stashed_resptr_props


def _convert_ark_v0_to_resource_iri(ark: str) -> str:
    """
    Converts an ARK URL from salsah.org (ARK version 0) of the form ark:/72163/080c-779b9990a0c3f-6e to a DSP resource
    IRI of the form http://rdfh.ch/080C/Ef9heHjPWDS7dMR_gGax2Q

    This method is needed for the migration of projects from salsah.org to DSP. Resources need to be created with an
    existing ARK, so the IRI needs to be extracted from that ARK in order for the ARK URL to be still valid after the
    migration.

    Args:
        ark: an ARK version 0 of the form ark:/72163/080c-779b9990a0c3f-6e, '72163' being the Name Assigning Authority
        number, '080c' being the project shortcode, '779b9990a0c3f' being an ID derived from the object's Salsah ID and
        '6e' being check digits

    Returns:
        Resource IRI (str) of the form http://rdfh.ch/080C/Ef9heHjPWDS7dMR_gGax2Q
    """
    # create the DaSCH namespace to create version 5 UUIDs
    generic_namespace_url = uuid.NAMESPACE_URL
    dasch_uuid_ns = uuid.uuid5(generic_namespace_url, "https://dasch.swiss")  # cace8b00-717e-50d5-bcb9-486f39d733a2

    # get the salsah resource ID from the ARK and convert it to a UUID version 5 (base64 encoded)
    if ark.count("-") != 2:
        raise BaseError(f"while converting ARK '{ark}'. The ARK seems to be invalid")
    project_id, resource_id, _ = ark.split("-")
    _, project_id = project_id.rsplit("/", 1)
    project_id = project_id.upper()
    if not re.match("^[0-9a-fA-F]{4}$", project_id):
        raise BaseError(f"while converting ARK '{ark}'. Invalid project shortcode '{project_id}'")
    if not re.match("^[0-9A-Za-z]+$", resource_id):
        raise BaseError(f"while converting ARK '{ark}'. Invalid Salsah ID '{resource_id}'")

    # make a UUID v5 from the namespace created above (which is a UUID itself) and the resource ID and encode it to base64
    dsp_uuid = base64.urlsafe_b64encode(uuid.uuid5(dasch_uuid_ns, resource_id).bytes).decode("utf-8")
    dsp_uuid = dsp_uuid[:-2]

    # use the new UUID to create the resource IRI
    return "http://rdfh.ch/" + project_id + "/" + dsp_uuid


def _parse_xml_file(input_file: str) -> etree.ElementTree:
    """
    Parse an XML file with DSP-conform data, remove namespace URI from the elements' names, and transform the special
    tags <annotation>, <region>, and <link> to their technically correct form <resource restype="Annotation">,
    <resource restype="Region">, and <resource restype="LinkObj">.

    Args:
        input_file: path to the XML file

    Returns:
        the parsed etree.ElementTree
    """
    tree = etree.parse(input_file)
    for elem in tree.getiterator():
        if not (isinstance(elem, etree._Comment) or isinstance(elem, etree._ProcessingInstruction)):
            # remove namespace URI in the element's name
            elem.tag = etree.QName(elem).localname
        if elem.tag == "annotation":
            elem.attrib["restype"] = "Annotation"
            elem.tag = "resource"
        elif elem.tag == "link":
            elem.attrib["restype"] = "LinkObj"
            elem.tag = "resource"
        elif elem.tag == "region":
            elem.attrib["restype"] = "Region"
            elem.tag = "resource"

    # remove unused namespace declarations
    etree.cleanup_namespaces(tree)

    return tree


def _check_consistency_with_ontology(
    resources: list[XMLResource],
    resclass_name_2_type: dict[str, type],
    verbose: bool = False
) -> None:
    """
    Checks if the resource types and properties in the XML are consistent with the ontology.

    Args:
        resources: a list of parsed XMLResources
        resclass_name_2_type: infos about the resource classes that exist on the DSP server for the current ontology
        verbose: verbose switch

    Returns:
        None if everything went well. Raises a BaseError if there is a problem.
    """
    if verbose:
        print("Check if the resource types and properties are consistent with the ontology...")
    knora_properties = resclass_name_2_type[resources[0].restype].knora_properties  # type: ignore

    for resource in resources:

        # check that the resource type is consistent with the ontology
        if resource.restype not in resclass_name_2_type:
            raise BaseError(
                f"=========================\n"
                f"ERROR: Resource '{resource.label}' (ID: {resource.id}) has an invalid resource type "
                f"'{resource.restype}'. Is your syntax correct? Remember the rules:\n"
                f" - DSP-API internals: <resource restype=\"restype\">         "
                        f"(will be interpreted as 'knora-api:restype')\n"
                f" - current ontology:  <resource restype=\":restype\">        "
                        f"('restype' must be defined in the 'resources' section of your ontology)\n"
                f" - other ontology:    <resource restype=\"other:restype\">   "
                        f"(not yet implemented: 'other' must be defined in the same JSON project file than your ontology)"
            )

        # check that the property types are consistent with the ontology
        resource_properties = resclass_name_2_type[resource.restype].properties.keys()  # type: ignore
        for propname in [prop.name for prop in resource.properties]:
            if propname not in knora_properties and propname not in resource_properties:
                raise BaseError(
                    f"=========================\n"
                    f"ERROR: Resource '{resource.label}' (ID: {resource.id}) has an invalid property '{propname}'. "
                    f"Is your syntax correct? Remember the rules:\n"
                    f" - DSP-API internals: <text-prop name=\"propname\">         "
                            f"(will be interpreted as 'knora-api:propname')\n"
                    f" - current ontology:  <text-prop name=\":propname\">        "
                            f"('propname' must be defined in the 'properties' section of your ontology)\n"
                    f" - other ontology:    <text-prop name=\"other:propname\">   "
                            f"(not yet implemented: 'other' must be defined in the same JSON project file than your ontology)"
                )

    print("Resource types and properties are consistent with the ontology.")


def xml_upload(input_file: str, server: str, user: str, password: str, imgdir: str, sipi: str, verbose: bool,
               incremental: bool, save_metrics: bool) -> bool:
    """
    This function reads an XML file and imports the data described in it onto the DSP server.

    Args:
        input_file: the XML with the data to be imported onto the DSP server
        server: the DSP server where the data should be imported
        user: the user (e-mail) with which the data should be imported
        password: the password of the user with which the data should be imported
        imgdir: the image directory
        sipi: the sipi instance to be used
        verbose: verbose option for the command, if used more output is given to the user
        incremental: if set, IRIs instead of internal IDs are expected as resource pointers
        save_metrics: if true, saves time measurements into a "metrics" folder in the current working directory

    Returns:
        True if all resources could be uploaded without errors; False if any resource (or part of it) could not be
        successfully uploaded
    """

    metrics: list[MetricRecord] = []
    preparation_start = datetime.now()

    # Validate the input XML file
    try:
        validate_xml_against_schema(input_file)
    except BaseError as err:
        print(f"=====================================\n"
              f"{err.message}")
        quit(0)

    save_location = Path.home() / Path(".dsp-tools")
    server_as_foldername = server
    server_substitutions = {
        r"https?://": "",
        r"^api\..+": "",
        r":\d{4}/?$": "",
        r"0.0.0.0": "localhost"
    }
    for pattern, repl in server_substitutions.items():
        server_as_foldername = re.sub(pattern, repl, server_as_foldername)

    # Connect to the DaSCH Service Platform API and get the project context
    con = Connection(server)
    try_network_action(failure_msg="Unable to login to DSP server", action=lambda: con.login(user, password))
    proj_context = try_network_action(failure_msg="Unable to retrieve project context from DSP server",
                                      action=lambda: ProjectContext(con=con))
    sipi_server = Sipi(sipi, con.get_token())

    # parse the XML file
    tree = _parse_xml_file(input_file)
    root = tree.getroot()
    default_ontology = root.attrib['default-ontology']
    shortcode = root.attrib['shortcode']
    resources: list[XMLResource] = []
    permissions: dict[str, XmlPermission] = {}
    for child in root:
        if child.tag == "permissions":
            permission = XmlPermission(child, proj_context)
            permissions[permission.id] = permission
        elif child.tag == "resource":
            resources.append(XMLResource(child, default_ontology))

    # get the project information and project ontology from the server
    res_inst_factory = try_network_action("", lambda: ResourceInstanceFactory(con, shortcode))
    permissions_lookup: dict[str, Permissions] = {s: perm.get_permission_instance() for s, perm in permissions.items()}
    resclass_name_2_type: dict[str, type] = {s: res_inst_factory.get_resclass_type(s) for s in res_inst_factory.get_resclass_names()}

    # check if the data in the XML is consistent with the ontology
    _check_consistency_with_ontology(
        resources=resources,
        resclass_name_2_type=resclass_name_2_type,
        verbose=verbose
    )

    # temporarily remove circular references, but only if not an incremental upload
    if not incremental:
        resources, stashed_xml_texts, stashed_resptr_props = _remove_circular_references(resources, verbose)
    else:
        stashed_xml_texts = dict()
        stashed_resptr_props = dict()

    preparation_duration = datetime.now() - preparation_start
    preparation_duration_ms = preparation_duration.seconds * 1000 + int(preparation_duration.microseconds / 1000)
    metrics.append(MetricRecord("", "", "", "xml upload preparation", preparation_duration_ms, ""))

    # upload all resources
    id2iri_mapping: dict[str, str] = {}
    failed_uploads: list[str] = []
    try:
        id2iri_mapping, failed_uploads, metrics = _upload_resources(
            resources, imgdir, sipi_server, permissions_lookup, resclass_name_2_type, id2iri_mapping, con,
            failed_uploads, metrics
        )
    except BaseException as err:
        _handle_upload_error(
            err=err,
            id2iri_mapping=id2iri_mapping,
            failed_uploads=failed_uploads,
            stashed_xml_texts=stashed_xml_texts,
            stashed_resptr_props=stashed_resptr_props,
            proj_shortcode=shortcode,
            onto_name=default_ontology,
            server_as_foldername=server_as_foldername,
            save_location=save_location
        )

    # update the resources with the stashed XML texts
    if len(stashed_xml_texts) > 0:
        try:
            nonapplied_xml_texts = _upload_stashed_xml_texts(verbose, id2iri_mapping, con, stashed_xml_texts)
            if len(nonapplied_xml_texts) > 0:
                raise BaseError(f"Error while trying to upload the stashed xml texts")
        except BaseException as err:
            _handle_upload_error(
                err=err,
                id2iri_mapping=id2iri_mapping,
                failed_uploads=failed_uploads,
                stashed_xml_texts=stashed_xml_texts,
                stashed_resptr_props=stashed_resptr_props,
                proj_shortcode=shortcode,
                onto_name=default_ontology,
                server_as_foldername=server_as_foldername,
                save_location=save_location
            )

    # update the resources with the stashed resptrs
    if len(stashed_resptr_props) > 0:
        try:
            nonapplied_resptr_props = _upload_stashed_resptr_props(verbose, id2iri_mapping, con, stashed_resptr_props)
            if len(nonapplied_resptr_props) > 0:
                raise BaseError(f"Error while trying to upload the stashed resptr props")
        except BaseException as err:
            _handle_upload_error(
                err=err,
                id2iri_mapping=id2iri_mapping,
                failed_uploads=failed_uploads,
                stashed_xml_texts=stashed_xml_texts,
                stashed_resptr_props=stashed_resptr_props,
                proj_shortcode=shortcode,
                onto_name=default_ontology,
                server_as_foldername=server_as_foldername,
                save_location=save_location
            )

    # write log files
    success = True
    timestamp_str = datetime.now().strftime("%Y-%m-%d_%H%M%S")
    id2iri_mapping_file = f"id2iri_{Path(input_file).stem}_mapping_{timestamp_str}.json"
    with open(id2iri_mapping_file, "x") as f:
        json.dump(id2iri_mapping, f, ensure_ascii=False, indent=4)
        print(f"The mapping of internal IDs to IRIs was written to {id2iri_mapping_file}")
    if failed_uploads:
        print(f"\nWARNING: Could not upload the following resources: {failed_uploads}\n")
        success = False
    if save_metrics:
        os.makedirs("metrics", exist_ok=True)
        df = pd.DataFrame(metrics)
        df.to_csv(f"metrics/{timestamp_str}_metrics_{server_as_foldername}_{Path(input_file).stem}.csv")
    if success:
        print("All resources have successfully been uploaded.")

    return success


def _upload_resources(
    resources: list[XMLResource],
    imgdir: str,
    sipi_server: Sipi,
    permissions_lookup: dict[str, Permissions],
    resclass_name_2_type: dict[str, type],
    id2iri_mapping: dict[str, str],
    con: Connection,
    failed_uploads: list[str],
    metrics: list[MetricRecord]
) -> tuple[dict[str, str], list[str], list[MetricRecord]]:
    """
    Iterates through all resources and tries to upload them to DSP

    Args:
        resources: list of XMLResources to upload to DSP
        imgdir: folder containing the multimedia files
        sipi_server: Sipi instance
        permissions_lookup: maps permission strings to Permission objects
        resclass_name_2_type: maps resource class names to their types
        id2iri_mapping: mapping of ids from the XML file to IRIs in DSP (initially empty, gets filled during the upload)
        con: connection to DSP
        failed_uploads: ids of resources that could not be uploaded (initially empty, gets filled during the upload)
        metrics: list with the metric records collected until now (gets filled during the upload)

    Returns:
        id2iri_mapping, failed_uploads, metrics
    """

    # If there are multimedia files: calculate their total size
    bitstream_all_sizes_mb = [Path(Path(imgdir) / Path(res.bitstream.value)).stat().st_size / 1000000 if res.bitstream else 0.0 for res in resources]
    if sum(bitstream_all_sizes_mb) > 0:
        bitstream_size_total_mb = round(sum(bitstream_all_sizes_mb), 1)
        bitstream_size_uploaded_mb = 0.0
        print(f"This xmlupload contains multimedia files with a total size of {bitstream_size_total_mb} MB.")

    for i, resource in enumerate(resources):
        resource_start = datetime.now()
        filetype = ""
        filesize = round(bitstream_all_sizes_mb[i], 1)
        bitstream_duration_ms = None
        resource_iri = resource.iri
        if resource.ark:
            resource_iri = _convert_ark_v0_to_resource_iri(resource.ark)

        # in case of a multimedia resource: upload the multimedia file
        resource_bitstream = None
        if resource.bitstream:
            try:
                bitstream_start = datetime.now()
                filetype = Path(resource.bitstream.value).suffix[1:]
                img: Optional[dict[Any, Any]] = try_network_action(
                    action=lambda: sipi_server.upload_bitstream(filepath=str(Path(imgdir) / Path(resource.bitstream.value))),  # type: ignore
                    failure_msg=f'ERROR while trying to upload file "{resource.bitstream.value}" of resource '
                                f'"{resource.label}" ({resource.id}).'
                )
                bitstream_duration = datetime.now() - bitstream_start
                bitstream_duration_ms = bitstream_duration.seconds * 1000 + int(bitstream_duration.microseconds / 1000)
                mb_per_sec = round((filesize / bitstream_duration_ms) * 1000, 1)
                metrics.append(MetricRecord(resource.id, filetype, filesize, "bitstream upload", bitstream_duration_ms, mb_per_sec))
            except BaseError as err:
                print(err.message)
                failed_uploads.append(resource.id)
                continue
            bitstream_size_uploaded_mb += bitstream_all_sizes_mb[i]
            print(f"Uploaded file '{resource.bitstream.value}' ({bitstream_size_uploaded_mb:.1f} MB / {bitstream_size_total_mb} MB)")
            internal_file_name_bitstream = img['uploadedFiles'][0]['internalFilename']  # type: ignore
            resource_bitstream = resource.get_bitstream(internal_file_name_bitstream, permissions_lookup)

        # create the resource in DSP
        resclass_type = resclass_name_2_type[resource.restype]
        properties = resource.get_propvals(id2iri_mapping, permissions_lookup)
        try:
            resource_instance: ResourceInstance = resclass_type(
                con=con,
                label=resource.label,
                iri=resource_iri,
                permissions=permissions_lookup.get(resource.permissions),  # type: ignore
                creation_date=resource.creation_date,
                bitstream=resource_bitstream,
                values=properties
            )
            resource_creation_start = datetime.now()
            created_resource: ResourceInstance = try_network_action(
                action=lambda: resource_instance.create(),
                failure_msg=f"ERROR while trying to create resource '{resource.label}' ({resource.id})."
            )
            resource_creation_duration = datetime.now() - resource_creation_start
            resource_creation_duration_ms = resource_creation_duration.seconds * 1000 + int(resource_creation_duration.microseconds / 1000)
            metrics.append(MetricRecord(resource.id, filetype, filesize, "resource creation", resource_creation_duration_ms, ""))
        except BaseError as err:
            print(err.message)
            failed_uploads.append(resource.id)
            continue
        id2iri_mapping[resource.id] = created_resource.iri
        print(f"Created resource {i+1}/{len(resources)}: '{created_resource.label}' (ID: '{resource.id}', IRI: "
              f"'{created_resource.iri}')")

        resource_duration = datetime.now() - resource_start
        resource_duration_ms = resource_duration.seconds * 1000 + int(resource_duration.microseconds / 1000)
        looping_overhead_ms = resource_duration_ms - resource_creation_duration_ms - (bitstream_duration_ms or 0)
        metrics.append(MetricRecord(resource.id, filetype, filesize, "looping overhead", looping_overhead_ms, ""))

    return id2iri_mapping, failed_uploads, metrics


def _upload_stashed_xml_texts(
    verbose: bool,
    id2iri_mapping: dict[str, str],
    con: Connection,
    stashed_xml_texts: dict[XMLResource, dict[XMLProperty, dict[str, KnoraStandoffXml]]]
) -> dict[XMLResource, dict[XMLProperty, dict[str, KnoraStandoffXml]]]:
    """
    After all resources are uploaded, the stashed xml texts must be applied to their resources in DSP.

    Args:
        verbose: bool
        id2iri_mapping: mapping of ids from the XML file to IRIs in DSP
        con: connection to DSP
        stashed_xml_texts: all xml texts that have been stashed

    Returns:
        nonapplied_xml_texts: the xml texts that could not be uploaded
    """

    print('Upload the stashed XML texts...')
    for resource, link_props in stashed_xml_texts.copy().items():
        if resource.id not in id2iri_mapping:
            # resource could not be uploaded to DSP, so the stash cannot be uploaded either
            continue
        res_iri = id2iri_mapping[resource.id]
        try:
            existing_resource = try_network_action(
                action=lambda: con.get(path=f'/v2/resources/{quote_plus(res_iri)}'),
                failure_msg=f'  ERROR while retrieving resource "{resource.id}" from DSP server.'
            )
        except BaseError as err:
            print(err.message)
            continue
        print(f'  Upload XML text(s) of resource "{resource.id}"...')
        for link_prop, hash_to_value in link_props.items():
            existing_values = existing_resource[link_prop.name]
            if not isinstance(existing_values, list):
                existing_values = [existing_values, ]
            for existing_value in existing_values:
                old_xmltext = existing_value.get("knora-api:textValueAsXml")
                if not old_xmltext:
                    continue

                # strip all xml tags from the old xmltext, so that the pure text itself remains
                pure_text = re.sub(r'(<\?xml.+>\s*)?<text>\s*(.+)\s*<\/text>', r'\2', old_xmltext)

                # if the pure text is a hash, the replacement must be made. This hash originates from
                # _stash_circular_references(), and identifies the XML texts
                if pure_text not in hash_to_value:
                    continue
                new_xmltext = hash_to_value[pure_text]

                # replace the outdated internal ids by their IRI
                for _id, _iri in id2iri_mapping.items():
                    new_xmltext.regex_replace(f'href="IRI:{_id}:IRI"', f'href="{_iri}"')

                # prepare API call
                jsonobj = {
                    "@id": res_iri,
                    "@type": resource.restype,
                    link_prop.name: {
                        "@id": existing_value['@id'],
                        "@type": "knora-api:TextValue",
                        "knora-api:textValueAsXml": new_xmltext,
                        "knora-api:textValueHasMapping": {
                            '@id': 'http://rdfh.ch/standoff/mappings/StandardMapping'
                        }
                    },
                    "@context": existing_resource['@context']
                }
                jsondata = json.dumps(jsonobj, indent=4, separators=(',', ': '), cls=KnoraStandoffXmlEncoder)

                # execute API call
                try:
                    try_network_action(
                        action=lambda: con.put(path='/v2/values', jsondata=jsondata),
                        failure_msg=f'    ERROR while uploading the xml text of "{link_prop.name}" of resource "{resource.id}"'
                    )
                except BaseError as err:
                    print(err.message)
                    continue
                stashed_xml_texts[resource][link_prop].pop(pure_text)
                if verbose:
                    print(f'  Successfully uploaded xml text of "{link_prop.name}"\n')

    # make a purged version of stashed_xml_texts, without empty entries
    nonapplied_xml_texts = _purge_stashed_xml_texts(stashed_xml_texts, id2iri_mapping)
    return nonapplied_xml_texts


def _purge_stashed_xml_texts(
    stashed_xml_texts: dict[XMLResource, dict[XMLProperty, dict[str, KnoraStandoffXml]]],
    id2iri_mapping: dict[str, str]
) -> dict[XMLResource, dict[XMLProperty, dict[str, KnoraStandoffXml]]]:
    """
    Accepts a stash of XML texts and purges it of resources that could not be uploaded (=don't exist in DSP), and of
    resources that had all their XML texts reapplied. It returns a new dict with only the resources that exist in DSP,
    but whose XML texts could not all be uploaded.

    Args:
        stashed_xml_texts: the stash to purge
        id2iri_mapping: used to check if a resource could be uploaded

    Returns:
        a purged version of stashed_xml_text
    """
    # remove resources that couldn't be uploaded. If they don't exist in DSP, it's not worth caring about their xmltexts
    stashed_xml_texts = {res: propdict for res, propdict in stashed_xml_texts.items() if res.id in id2iri_mapping}

    # remove resources that don't have stashed xmltexts (=all xmltexts had been reapplied)
    nonapplied_xml_texts: dict[XMLResource, dict[XMLProperty, dict[str, KnoraStandoffXml]]] = {}
    for res, propdict in stashed_xml_texts.items():
        for prop, xmldict in propdict.items():
            if len(xmldict) > 0:
                if res not in nonapplied_xml_texts:
                    nonapplied_xml_texts[res] = {}
                nonapplied_xml_texts[res][prop] = xmldict
    return nonapplied_xml_texts


def _upload_stashed_resptr_props(
    verbose: bool,
    id2iri_mapping: dict[str, str],
    con: Connection,
    stashed_resptr_props: dict[XMLResource, dict[XMLProperty, list[str]]]
) -> dict[XMLResource, dict[XMLProperty, list[str]]]:
    """
    After all resources are uploaded, the stashed resptr props must be applied to their resources in DSP.

    Args:
        verbose: bool
        id2iri_mapping: mapping of ids from the XML file to IRIs in DSP
        con: connection to DSP
        stashed_resptr_props: all resptr props that have been stashed

    Returns:
        nonapplied_resptr_props: the resptr props that could not be uploaded
    """

    print('Upload the stashed resptrs...')
    for resource, prop_2_resptrs in stashed_resptr_props.copy().items():
        if resource.id not in id2iri_mapping:
            # resource could not be uploaded to DSP, so the stash cannot be uploaded either
            continue
        res_iri = id2iri_mapping[resource.id]
        try:
            existing_resource = try_network_action(
                action=lambda: con.get(path=f'/v2/resources/{quote_plus(res_iri)}'),
                failure_msg=f'  ERROR while retrieving resource "{resource.id}" from DSP server'
            )
        except BaseError as err:
            print(err.message)
            continue
        print(f'  Upload resptrs of resource "{resource.id}"...')
        for link_prop, resptrs in prop_2_resptrs.items():
            for resptr in resptrs.copy():
                jsonobj = {
                    '@id': res_iri,
                    '@type': resource.restype,
                    f'{link_prop.name}Value': {
                        '@type': 'knora-api:LinkValue',
                        'knora-api:linkValueHasTargetIri': {
                            # if target doesn't exist in DSP, send the (invalid) resource ID of target to DSP, which
                            # will produce an understandable error message
                            '@id': id2iri_mapping.get(resptr, resptr)
                        }
                    },
                    '@context': existing_resource['@context']
                }
                jsondata = json.dumps(jsonobj, indent=4, separators=(',', ': '))
                try:
                    try_network_action(
                        action=lambda: con.post(path='/v2/values', jsondata=jsondata),
                        failure_msg=f'    ERROR while uploading the resptr prop of "{link_prop.name}" of resource "{resource.id}"'
                    )
                except BaseError as err:
                    print(err.message)
                    continue
                stashed_resptr_props[resource][link_prop].remove(resptr)
                if verbose:
                    print(f'  Successfully uploaded resptr-prop of "{link_prop.name}"\n'
                          f'    Value: {resptr}')

    # make a purged version of stashed_resptr_props, without empty entries
    nonapplied_resptr_props = _purge_stashed_resptr_props(stashed_resptr_props, id2iri_mapping)
    return nonapplied_resptr_props


def _purge_stashed_resptr_props(
    stashed_resptr_props: dict[XMLResource, dict[XMLProperty, list[str]]],
    id2iri_mapping: dict[str, str]
) -> dict[XMLResource, dict[XMLProperty, list[str]]]:
    """
    Accepts a stash of resptrs and purges it of resources that could not be uploaded (=don't exist in DSP), and of
    resources that had all their resptrs reapplied. It returns a new dict with only the resources that exist in DSP,
    but whose resptrs could not all be uploaded.

    Args:
        stashed_resptr_props: the stash to purge
        id2iri_mapping: used to check if a resource could be uploaded (optional)

    Returns:
        a purged version of stashed_resptr_props
    """
    # remove resources that couldn't be uploaded. If they don't exist in DSP, it's not worth caring about their resptrs
    stashed_resptr_props = {res: pdict for res, pdict in stashed_resptr_props.items() if res.id in id2iri_mapping}

    # remove resources that don't have stashed resptrs (=all resptrs had been reapplied)
    nonapplied_resptr_props: dict[XMLResource, dict[XMLProperty, list[str]]] = {}
    for res, propdict in stashed_resptr_props.items():
        for prop, resptrs in propdict.items():
            if len(resptrs) > 0:
                if res not in nonapplied_resptr_props:
                    nonapplied_resptr_props[res] = {}
                nonapplied_resptr_props[res][prop] = resptrs
    return nonapplied_resptr_props


def _handle_upload_error(
    err: BaseException,
    id2iri_mapping: dict[str, str],
    failed_uploads: list[str],
    stashed_xml_texts: dict[XMLResource, dict[XMLProperty, dict[str, KnoraStandoffXml]]],
    stashed_resptr_props: dict[XMLResource, dict[XMLProperty, list[str]]],
    proj_shortcode: str,
    onto_name: str,
    server_as_foldername: str,
    save_location: Path
) -> None:
    """
    In case the xmlupload must be interrupted, e.g. because of an error that could not be handled, or due to keyboard
    interrupt, this method ensures that all information about what is already in DSP is written into log files.

    It then re-raises the original error.

    Args:
        err: error that was the cause of the abort
        id2iri_mapping: mapping of ids from the XML file to IRIs in DSP (only successful uploads appear here)
        failed_uploads: resources that caused an error when uploading to DSP
        stashed_xml_texts: all xml texts that have been stashed
        stashed_resptr_props: all resptr props that have been stashed
        proj_shortcode: shortcode of the project the data belongs to
        onto_name: name of the ontology the data references
        server_as_foldername: the server which the data is uploaded onto (in a form that can be used as folder name)
        save_location: path where to save the logs

    Returns:
        None
    """

    print(f'\n=========================================='
          f'\nxmlupload must be aborted because of an error')
    timestamp_str = datetime.now().strftime("%Y-%m-%d_%H%M%S")

    save_location_full = save_location / "xmluploads" / server_as_foldername / proj_shortcode / onto_name
    save_location_full.mkdir(parents=True, exist_ok=True)

    # only stashed properties of resources that already exist in DSP are of interest
    stashed_xml_texts = _purge_stashed_xml_texts(stashed_xml_texts, id2iri_mapping)
    stashed_resptr_props = _purge_stashed_resptr_props(stashed_resptr_props, id2iri_mapping)

    if id2iri_mapping:
        id2iri_mapping_file = f"{save_location_full}/{timestamp_str}_id2iri_mapping.json"
        with open(id2iri_mapping_file, "x") as f:
            json.dump(id2iri_mapping, f, ensure_ascii=False, indent=4)
        print(f"The mapping of internal IDs to IRIs was written to {id2iri_mapping_file}")

    if stashed_xml_texts:
        stashed_xml_texts_serializable = {r.id: {p.name: xml for p, xml in rdict.items()} for r, rdict in stashed_xml_texts.items()}
        xml_filename = f"{save_location_full}/{timestamp_str}_stashed_text_properties.json"
        with open(xml_filename, "x") as f:
            json.dump(stashed_xml_texts_serializable, f, ensure_ascii=False, indent=4, cls=KnoraStandoffXmlEncoder)
        print(f"There are stashed text properties that could not be reapplied to the resources they were stripped "
              f"from. They were saved to {xml_filename}.")

    if stashed_resptr_props:
        stashed_resptr_props_serializable = {r.id: {p.name: plist for p, plist in rdict.items()} for r, rdict in stashed_resptr_props.items()}
        resptr_filename = f"{save_location_full}/{timestamp_str}_stashed_resptr_properties.json"
        with open(resptr_filename, "x") as f:
            json.dump(stashed_resptr_props_serializable, f, ensure_ascii=False, indent=4)
        print(
            f"There are stashed resptr properties that could not be reapplied to the resources they were stripped "
            f"from. They were saved to {resptr_filename}")

    # print the resources that threw an error when they were tried to be uploaded
    if failed_uploads:
        print(f"Independently of this error, there were some resources that could not be uploaded: "
              f"{failed_uploads}")

    if isinstance(err, KeyboardInterrupt):
        exit(1)
    else:
        print('The error will now be raised again:\n'
              '==========================================\n')
        raise err