Before: `./sitename`
After: `sitename`
This used to return `.` before, even though these are paths identical
they aren't REALLY. Tar command strips 2 subpaths so this ended up
breaking file restores.
* Revert "chore: move function to correct file"
This reverts commit ebfdfa283b.
* Revert "refactor!: merge get_site_url into get_url (#22308)"
This reverts commit 2001bc278f.
- Move the config to bench level and not site level because, server
script "threat model" requires consent from a bench owner and not
individual site.
- While this is a breaking change which people may not like, we believe
it's essential to improve security model of Frappe.
We are dropping some keys which are hardcoded in function.
- this over time might be out of sync with rest of the code.
- other apps cache stuff too.
Best fix: just clear everything.
The parent folder of a python module (inside `./site-packages`) can be
`./lib`. Separate code paths that depend on the app source into its own
getter so that patching that function becomes easier in those cases
where the python module is not served from the source checkout.
* fix: Remove re cache internals manipulation
* fix: Purge re cache after module loads
Empty cache would work better as we already got our pre-compiled
patterns at the top level of every module. This leaves the cache open
for dynamically generated patterns which are in better need of it. Over
time, workers would converge to this anyway. This change only reduces
the cache hit and eviction effort.
I'd improve this by executing `re.purge` on every module import but
complexity tradeoff lol. I'd prefer if re didn't cache patterns
generated by `re.compile` but I dont see this behaviour or any escape
hatches so this will have to do for now.
* fix: procure db config from single authority
ensures that configuration is uniformely procured from local.conf
instead of making use of hard to audit multilevel fallback logic
Implementation Note:
- `get_db(host, port, user, password)` was stripped of any optional
argument and therefrom all errors where fixed.
- All occurances of `grep 'frappe.db.db_'` where changed to
`frappe.conf.db_`
* fix: revert unnecessary breaking changes
This avoids having to manipulate config files in brittle bash
entrypoints that need to react to dynamic service discovery.
This significantly improves the operability of various bench sites.
This is running on several prod site without any noticable problems, so
making it default behaviour.
Opt out by setting env variable `FRAPPE_TUNE_GC=False`
This is more intuitive and consistent with other things like `frappe.db`.
PS: This is quite likely to break some weird usage which I can't guess right now. Normal usage inside request/job cycles will continue to work as it used to.
Each call to evaluate if route is web view makes N queries where N = #
of web view doctypes. This entire computation can be definitely cached
for short duration.
- Added cache bursting in WebsiteGenerator doctype updates.
- Added 60 minutes TTL in case cache invalidation wasn't done reliably.
- Hashes are supposed to be used for representing complex object, not
multiple documents of same DocType.
- Redis's auto cache clearing wont clear individual key from hashes even if they
are rarely used.
- Keys can have expiry.
Passing lambda function from inside document object would keep reference
to document alive. This means increasing memeory usage in bulk
processing.
Refer https://github.com/frappe/frappe/pull/17061 for example
This also extends it to db.set_value
