reporting-engine/report_xlsx/controllers/main.py

108 lines
4.3 KiB
Python
Raw Permalink Normal View History

2024-05-03 09:07:13 +00:00
# Copyright (C) 2017 Creu Blanca
# Copyright 2021 ForgeFlow S.L.
# License AGPL-3.0 or later (https://www.gnuorg/licenses/agpl.html).
import json
import logging
from werkzeug.urls import url_decode
from flectra.http import (
content_disposition,
request,
route,
)
from flectra.http import (
serialize_exception as _serialize_exception,
)
from flectra.tools import html_escape
from flectra.tools.safe_eval import safe_eval, time
from flectra.addons.web.controllers.report import ReportController
_logger = logging.getLogger(__name__)
class ReportController(ReportController):
@route()
def report_routes(self, reportname, docids=None, converter=None, **data):
if converter == "xlsx":
report = request.env["ir.actions.report"]._get_report_from_name(reportname)
context = dict(request.env.context)
if docids:
docids = [int(i) for i in docids.split(",")]
if data.get("options"):
data.update(json.loads(data.pop("options")))
if data.get("context"):
data["context"] = json.loads(data["context"])
context.update(data["context"])
xlsx = report.with_context(**context)._render_xlsx(
reportname, docids, data=data
)[0]
xlsxhttpheaders = [
(
"Content-Type",
"application/vnd.openxmlformats-"
"officedocument.spreadsheetml.sheet",
),
("Content-Length", len(xlsx)),
]
return request.make_response(xlsx, headers=xlsxhttpheaders)
return super().report_routes(reportname, docids, converter, **data)
@route()
def report_download(self, data, context=None, token=None):
requestcontent = json.loads(data)
url, report_type = requestcontent[0], requestcontent[1]
if report_type == "xlsx":
try:
reportname = url.split("/report/xlsx/")[1].split("?")[0]
docids = None
if "/" in reportname:
reportname, docids = reportname.split("/")
if docids:
# Generic report:
response = self.report_routes(
reportname, docids=docids, converter="xlsx", context=context
)
else:
# Particular report:
data = dict(
url_decode(url.split("?")[1]).items()
) # decoding the args represented in JSON
if "context" in data:
context, data_context = (
json.loads(context or "{}"),
json.loads(data.pop("context")),
)
context = json.dumps({**context, **data_context})
response = self.report_routes(
reportname, converter="xlsx", context=context, **data
)
report = request.env["ir.actions.report"]._get_report_from_name(
reportname
)
filename = f"{report.name}.xlsx"
if docids:
ids = [int(x) for x in docids.split(",")]
obj = request.env[report.model].browse(ids)
if report.print_report_name and not len(obj) > 1:
report_name = safe_eval(
report.print_report_name, {"object": obj, "time": time}
)
filename = f"{report_name}.xlsx"
if not response.headers.get("Content-Disposition"):
response.headers.add(
"Content-Disposition", content_disposition(filename)
)
return response
except Exception as e:
_logger.exception("Error while generating report %s", reportname)
se = _serialize_exception(e)
error = {"code": 200, "message": "Flectra Server Error", "data": se}
return request.make_response(html_escape(json.dumps(error)))
else:
return super().report_download(data, context=context, token=token)