Optimize format_elapsed_time

The optimized code achieves a **147% speedup** (from 1.92ms to 775μs) by eliminating expensive `timedelta` object creation and operations, replacing them with simple float arithmetic.

**Key optimizations:**

1. **Removed timedelta overhead**: The original code created multiple `timedelta` objects for comparisons and calculations. Line profiler shows `timedelta(seconds=elapsed_time)` alone consumed 20.6% of runtime. The optimized version uses direct float comparisons with pre-defined constants (`_ONE_SECOND`, `_ONE_MINUTE`).

2. **Simplified milliseconds calculation**: Changed from `delta / timedelta(milliseconds=1)` to `elapsed_time * _MS_PER_SECOND`. This eliminates timedelta division overhead and uses a simple multiplication.

3. **Direct arithmetic for minutes/seconds**: Replaced `delta // timedelta(minutes=1)` and `(delta - timedelta(minutes=minutes)).total_seconds()` with straightforward integer division and subtraction (`elapsed_time // _ONE_MINUTE` and `elapsed_time - minutes * _ONE_MINUTE`).

**Why it's faster:**

- **Object creation cost**: Each `timedelta()` call involves object allocation and initialization. The original creates 3-5 timedelta objects per invocation depending on the code path.
- **Simpler operations**: Float comparisons and arithmetic are primitive CPU operations, while timedelta comparisons/operations involve method calls and attribute access.
- **Line profiler evidence**: The initial `delta = timedelta(seconds=elapsed_time)` took 878μs (20.6%), while the optimized first comparison takes only 288μs (12.2%).

**Test case performance:**

The optimization benefits all test cases uniformly since it improves the fundamental operations. Tests show correctness is preserved across edge cases (boundary conditions, rounding, singular/plural forms) while delivering consistent speedup for milliseconds (< 1s), seconds (1-60s), and minutes+seconds (≥ 60s) ranges.

This is a pure performance win with no functional changes—ideal for merge.

Author: codeflash-ai[bot]

src/backend/base/langflow/api/utils/core.py

@@ -2,7 +2,6 @@
 
 import uuid
 from ast import literal_eval
-from datetime import timedelta
 from enum import Enum
 from typing import TYPE_CHECKING, Annotated, Any
 
@@ -28,6 +27,12 @@
     from langflow.services.chat.service import ChatService
     from langflow.services.store.schema import StoreComponentCreate
 
+_ONE_SECOND = 1.0
+
+_ONE_MINUTE = 60.0
+
+_MS_PER_SECOND = 1000.0
+
 
 API_WORDS = ["api", "key", "token"]
 
@@ -153,18 +158,17 @@ def format_elapsed_time(elapsed_time: float) -> str:
     - Less than 1 minute: returns seconds rounded to 1 decimal
     - 1 minute or more: returns minutes and seconds
     """
-    delta = timedelta(seconds=elapsed_time)
-    if delta < timedelta(seconds=1):
-        milliseconds = round(delta / timedelta(milliseconds=1))
+    if elapsed_time < _ONE_SECOND:
+        milliseconds = round(elapsed_time * _MS_PER_SECOND)
         return f"{milliseconds} ms"
 
-    if delta < timedelta(minutes=1):
+    if elapsed_time < _ONE_MINUTE:
         seconds = round(elapsed_time, 1)
         unit = "second" if seconds == 1 else "seconds"
         return f"{seconds} {unit}"
 
-    minutes = delta // timedelta(minutes=1)
-    seconds = round((delta - timedelta(minutes=minutes)).total_seconds(), 1)
+    minutes = int(elapsed_time // _ONE_MINUTE)
+    seconds = round(elapsed_time - minutes * _ONE_MINUTE, 1)
     minutes_unit = "minute" if minutes == 1 else "minutes"
     seconds_unit = "second" if seconds == 1 else "seconds"
     return f"{minutes} {minutes_unit}, {seconds} {seconds_unit}"