mirror of
https://github.com/bookwyrm-social/bookwyrm.git
synced 2024-10-31 22:19:00 +00:00
Merge branch 'main' into mypy-isbn
This commit is contained in:
commit
5eae123668
6 changed files with 62 additions and 28 deletions
|
@ -1,4 +1,6 @@
|
|||
""" handle reading a csv from calibre """
|
||||
from typing import Any, Optional
|
||||
|
||||
from bookwyrm.models import Shelf
|
||||
|
||||
from . import Importer
|
||||
|
@ -9,7 +11,7 @@ class CalibreImporter(Importer):
|
|||
|
||||
service = "Calibre"
|
||||
|
||||
def __init__(self, *args, **kwargs):
|
||||
def __init__(self, *args: Any, **kwargs: Any):
|
||||
# Add timestamp to row_mappings_guesses for date_added to avoid
|
||||
# integrity error
|
||||
row_mappings_guesses = []
|
||||
|
@ -23,6 +25,6 @@ class CalibreImporter(Importer):
|
|||
self.row_mappings_guesses = row_mappings_guesses
|
||||
super().__init__(*args, **kwargs)
|
||||
|
||||
def get_shelf(self, normalized_row):
|
||||
def get_shelf(self, normalized_row: dict[str, Optional[str]]) -> Optional[str]:
|
||||
# Calibre export does not indicate which shelf to use. Use a default one for now
|
||||
return Shelf.TO_READ
|
||||
|
|
|
@ -1,8 +1,10 @@
|
|||
""" handle reading a csv from an external service, defaults are from Goodreads """
|
||||
import csv
|
||||
from datetime import timedelta
|
||||
from typing import Iterable, Optional
|
||||
|
||||
from django.utils import timezone
|
||||
from bookwyrm.models import ImportJob, ImportItem, SiteSettings
|
||||
from bookwyrm.models import ImportJob, ImportItem, SiteSettings, User
|
||||
|
||||
|
||||
class Importer:
|
||||
|
@ -35,19 +37,26 @@ class Importer:
|
|||
}
|
||||
|
||||
# pylint: disable=too-many-locals
|
||||
def create_job(self, user, csv_file, include_reviews, privacy):
|
||||
def create_job(
|
||||
self, user: User, csv_file: Iterable[str], include_reviews: bool, privacy: str
|
||||
) -> ImportJob:
|
||||
"""check over a csv and creates a database entry for the job"""
|
||||
csv_reader = csv.DictReader(csv_file, delimiter=self.delimiter)
|
||||
rows = list(csv_reader)
|
||||
if len(rows) < 1:
|
||||
raise ValueError("CSV file is empty")
|
||||
rows = enumerate(rows)
|
||||
|
||||
mappings = (
|
||||
self.create_row_mappings(list(fieldnames))
|
||||
if (fieldnames := csv_reader.fieldnames)
|
||||
else {}
|
||||
)
|
||||
|
||||
job = ImportJob.objects.create(
|
||||
user=user,
|
||||
include_reviews=include_reviews,
|
||||
privacy=privacy,
|
||||
mappings=self.create_row_mappings(csv_reader.fieldnames),
|
||||
mappings=mappings,
|
||||
source=self.service,
|
||||
)
|
||||
|
||||
|
@ -55,16 +64,20 @@ class Importer:
|
|||
if enforce_limit and allowed_imports <= 0:
|
||||
job.complete_job()
|
||||
return job
|
||||
for index, entry in rows:
|
||||
for index, entry in enumerate(rows):
|
||||
if enforce_limit and index >= allowed_imports:
|
||||
break
|
||||
self.create_item(job, index, entry)
|
||||
return job
|
||||
|
||||
def update_legacy_job(self, job):
|
||||
def update_legacy_job(self, job: ImportJob) -> None:
|
||||
"""patch up a job that was in the old format"""
|
||||
items = job.items
|
||||
headers = list(items.first().data.keys())
|
||||
first_item = items.first()
|
||||
if first_item is None:
|
||||
return
|
||||
|
||||
headers = list(first_item.data.keys())
|
||||
job.mappings = self.create_row_mappings(headers)
|
||||
job.updated_date = timezone.now()
|
||||
job.save()
|
||||
|
@ -75,24 +88,24 @@ class Importer:
|
|||
item.normalized_data = normalized
|
||||
item.save()
|
||||
|
||||
def create_row_mappings(self, headers):
|
||||
def create_row_mappings(self, headers: list[str]) -> dict[str, Optional[str]]:
|
||||
"""guess what the headers mean"""
|
||||
mappings = {}
|
||||
for (key, guesses) in self.row_mappings_guesses:
|
||||
value = [h for h in headers if h.lower() in guesses]
|
||||
value = value[0] if len(value) else None
|
||||
values = [h for h in headers if h.lower() in guesses]
|
||||
value = values[0] if len(values) else None
|
||||
if value:
|
||||
headers.remove(value)
|
||||
mappings[key] = value
|
||||
return mappings
|
||||
|
||||
def create_item(self, job, index, data):
|
||||
def create_item(self, job: ImportJob, index: int, data: dict[str, str]) -> None:
|
||||
"""creates and saves an import item"""
|
||||
normalized = self.normalize_row(data, job.mappings)
|
||||
normalized["shelf"] = self.get_shelf(normalized)
|
||||
ImportItem(job=job, index=index, data=data, normalized_data=normalized).save()
|
||||
|
||||
def get_shelf(self, normalized_row):
|
||||
def get_shelf(self, normalized_row: dict[str, Optional[str]]) -> Optional[str]:
|
||||
"""determine which shelf to use"""
|
||||
shelf_name = normalized_row.get("shelf")
|
||||
if not shelf_name:
|
||||
|
@ -103,11 +116,15 @@ class Importer:
|
|||
]
|
||||
return shelf[0] if shelf else None
|
||||
|
||||
def normalize_row(self, entry, mappings): # pylint: disable=no-self-use
|
||||
# pylint: disable=no-self-use
|
||||
def normalize_row(
|
||||
self, entry: dict[str, str], mappings: dict[str, Optional[str]]
|
||||
) -> dict[str, Optional[str]]:
|
||||
"""use the dataclass to create the formatted row of data"""
|
||||
return {k: entry.get(v) for k, v in mappings.items()}
|
||||
return {k: entry.get(v) if v else None for k, v in mappings.items()}
|
||||
|
||||
def get_import_limit(self, user): # pylint: disable=no-self-use
|
||||
# pylint: disable=no-self-use
|
||||
def get_import_limit(self, user: User) -> tuple[int, int]:
|
||||
"""check if import limit is set and return how many imports are left"""
|
||||
site_settings = SiteSettings.objects.get()
|
||||
import_size_limit = site_settings.import_size_limit
|
||||
|
@ -125,7 +142,9 @@ class Importer:
|
|||
allowed_imports = import_size_limit - imported_books
|
||||
return enforce_limit, allowed_imports
|
||||
|
||||
def create_retry_job(self, user, original_job, items):
|
||||
def create_retry_job(
|
||||
self, user: User, original_job: ImportJob, items: list[ImportItem]
|
||||
) -> ImportJob:
|
||||
"""retry items that didn't import"""
|
||||
job = ImportJob.objects.create(
|
||||
user=user,
|
||||
|
|
|
@ -1,11 +1,16 @@
|
|||
""" handle reading a tsv from librarything """
|
||||
import re
|
||||
from typing import Optional
|
||||
|
||||
from bookwyrm.models import Shelf
|
||||
|
||||
from . import Importer
|
||||
|
||||
|
||||
def _remove_brackets(value: Optional[str]) -> Optional[str]:
|
||||
return re.sub(r"\[|\]", "", value) if value else None
|
||||
|
||||
|
||||
class LibrarythingImporter(Importer):
|
||||
"""csv downloads from librarything"""
|
||||
|
||||
|
@ -13,16 +18,19 @@ class LibrarythingImporter(Importer):
|
|||
delimiter = "\t"
|
||||
encoding = "ISO-8859-1"
|
||||
|
||||
def normalize_row(self, entry, mappings): # pylint: disable=no-self-use
|
||||
def normalize_row(
|
||||
self, entry: dict[str, str], mappings: dict[str, Optional[str]]
|
||||
) -> dict[str, Optional[str]]: # pylint: disable=no-self-use
|
||||
"""use the dataclass to create the formatted row of data"""
|
||||
remove_brackets = lambda v: re.sub(r"\[|\]", "", v) if v else None
|
||||
normalized = {k: remove_brackets(entry.get(v)) for k, v in mappings.items()}
|
||||
isbn_13 = normalized.get("isbn_13")
|
||||
isbn_13 = isbn_13.split(", ") if isbn_13 else []
|
||||
normalized = {
|
||||
k: _remove_brackets(entry.get(v) if v else None)
|
||||
for k, v in mappings.items()
|
||||
}
|
||||
isbn_13 = value.split(", ") if (value := normalized.get("isbn_13")) else []
|
||||
normalized["isbn_13"] = isbn_13[1] if len(isbn_13) > 1 else None
|
||||
return normalized
|
||||
|
||||
def get_shelf(self, normalized_row):
|
||||
def get_shelf(self, normalized_row: dict[str, Optional[str]]) -> Optional[str]:
|
||||
if normalized_row["date_finished"]:
|
||||
return Shelf.READ_FINISHED
|
||||
if normalized_row["date_started"]:
|
||||
|
|
|
@ -1,4 +1,6 @@
|
|||
""" handle reading a csv from openlibrary"""
|
||||
from typing import Any
|
||||
|
||||
from . import Importer
|
||||
|
||||
|
||||
|
@ -7,7 +9,7 @@ class OpenLibraryImporter(Importer):
|
|||
|
||||
service = "OpenLibrary"
|
||||
|
||||
def __init__(self, *args, **kwargs):
|
||||
def __init__(self, *args: Any, **kwargs: Any):
|
||||
self.row_mappings_guesses.append(("openlibrary_key", ["edition id"]))
|
||||
self.row_mappings_guesses.append(("openlibrary_work_key", ["work id"]))
|
||||
super().__init__(*args, **kwargs)
|
||||
|
|
|
@ -54,10 +54,10 @@ ImportStatuses = [
|
|||
class ImportJob(models.Model):
|
||||
"""entry for a specific request for book data import"""
|
||||
|
||||
user = models.ForeignKey(User, on_delete=models.CASCADE)
|
||||
user: User = models.ForeignKey(User, on_delete=models.CASCADE)
|
||||
created_date = models.DateTimeField(default=timezone.now)
|
||||
updated_date = models.DateTimeField(default=timezone.now)
|
||||
include_reviews = models.BooleanField(default=True)
|
||||
include_reviews: bool = models.BooleanField(default=True)
|
||||
mappings = models.JSONField()
|
||||
source = models.CharField(max_length=100)
|
||||
privacy = models.CharField(max_length=255, default="public", choices=PrivacyLevels)
|
||||
|
@ -76,7 +76,7 @@ class ImportJob(models.Model):
|
|||
|
||||
self.save(update_fields=["task_id"])
|
||||
|
||||
def complete_job(self):
|
||||
def complete_job(self) -> None:
|
||||
"""Report that the job has completed"""
|
||||
self.status = "complete"
|
||||
self.complete = True
|
||||
|
|
3
mypy.ini
3
mypy.ini
|
@ -13,6 +13,9 @@ implicit_reexport = True
|
|||
[mypy-bookwyrm.connectors.*]
|
||||
ignore_errors = False
|
||||
|
||||
[mypy-bookwyrm.importers.*]
|
||||
ignore_errors = False
|
||||
|
||||
[mypy-bookwyrm.isbn.*]
|
||||
ignore_errors = False
|
||||
|
||||
|
|
Loading…
Reference in a new issue