qjc1997 opened a new pull request, #4:
URL: https://github.com/apache/casbin-python-redis-adapter/pull/4
## What
`Adapter.load_policy` used `LLEN` followed by one `LINDEX` per rule:
```python
length = self.client.llen(self.key)
for i in range(length):
line = self.client.lindex(self.key, i)
```
This PR fetches the whole list with a single `LRANGE key 0 -1` instead.
Parsing is unchanged (`json.loads` → `CasbinRule` →
`persist.load_policy_line`), and nothing on the write path changes.
## Why
**Latency.** Each `LINDEX` is a separate client↔Redis round trip, so loading
N rules costs N+1 round trips. The Redis server-side work is negligible; the
time goes into waiting on the network. Measured against a managed Redis (~0.75
ms RTT) with ~2,950 rules:
| | time |
|---|---|
| `LLEN` + N × `LINDEX` (current) | 2.22 s |
| the same `LINDEX` calls pipelined (1 round trip) | 0.026 s |
| 1 × `LRANGE key 0 -1` (this PR) | 0.002 s |
`load_policy` runs when an enforcer is created and again every time policy
is reloaded (watcher callback, explicit reloads), so applications pay this cost
repeatedly.
**Consistency.** `LLEN` + `LINDEX` is not atomic. If another process removes
a rule (`LREM`) while the loop is running, the list shrinks, the last `LINDEX`
calls return `None`, and `json.loads(None)` raises a `TypeError`; inserts or
removals in the middle can also skip or duplicate rules. `LRANGE` is a single
command, so it reads a consistent snapshot of the list.
This also brings the Python adapter in line with the Go redis-adapter, whose
`LoadPolicy` already reads the list with a single `LRANGE`.
## Tests
Added `test_load_policy_single_round_trip`: it reloads the policy with
`lindex` patched to fail and checks that `lrange` is called exactly once with
`(key, 0, -1)` and that the loaded policy enforces correctly. It fails on
current `master` (`LINDEX used`) and passes with this change. The full suite
passes locally (10 tests), and `black --check` is clean.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
