Performance regression with 2020-resolver

[kveretennicov]

Environment

• pip version: 20.2.3
• Python version: 3.8.5
• OS: Ubuntu 16.04

Description
Running subsequent pip install for a set of requirements with --use-feature=2020-resolver takes 10 times longer than without --use-feature=2020-resolver.

Expected behavior
Similar performance in both cases.

How to Reproduce

1. Create and activate a virtual environment.
2. Create a requirements.txt:

click~=7.1
colorama~=0.4
enlighten~=1.6.1
networkx~=2.3
numpy>=1.17.0
numba>=0.50
orjson~=3.1
pandas~=1.0
pint~=0.13
pyarrow~=0.17
pydantic~=1.5
PyYAML~=5.1
sentry-sdk~=0.15
tabulate>=0.8.7
termcolor~=1.1

3. Install pip 20.2.3
4. Do initial installation of requirements:

$ pip install -r requirements.txt

5. Repeat the installation and time it

$ time pip install -r requirements.txt
real	0m0.558s
user	0m0.508s
sys	0m0.044s

6. Now repeat the installation with --use-feature=2020-resolver and time it

$ time pip install --use-feature=2020-resolver -r requirements.txt
real	0m6.304s
user	0m5.136s
sys	0m0.136s

With --use-feature=2020-resolver the repeated installation is consistently ~10x slower.

[pradyunsg]

Thanks for filing this issue! That's a useful test case to have. Please see #8664. :)