patched function.py

from types import CodeType, FunctionType
from dis import dis

def patch(fn, name, codestring):
    code = fn.__code__
    code = CodeType(
        code.co_argcount, code.co_posonlyargcount, code.co_kwonlyargcount,
        code.co_nlocals, code.co_stacksize, code.co_flags, codestring,
        code.co_consts, code.co_names, code.co_varnames,
        code.co_filename, name, code.co_firstlineno, code.co_lnotab,
        code.co_freevars,  code.co_cellvars,
    )
    return FunctionType(
        code, fn.__globals__, name, fn.__defaults__, fn.__closure__
    )

def only_dict(d, i):
    d[i] = 0
    return d[i]

def without_walrus(d, i):
    r = 0
    d[i] = r
    return r

def with_walrus(d, i):
    d[i] = (r := 0)
    return r

patched_walrus = patch(with_walrus, 'patched_walrus', bytes([100, 1, 4, 0, 124, 0, 124, 1, 60, 0, 83, 0]))

dis(patched_walrus)

plot.py

import numpy as np
import pandas as pd
from scipy import stats
import matplotlib.pyplot as plt
import seaborn as sns

sns.set(style='ticks')

df = pd.read_csv('idx_vs_walrus.csv')

fig, (ax1, ax2) = plt.subplots(
    ncols=2, sharey=True, figsize=(9, 4),
    gridspec_kw=dict(width_ratios=[10, 1]),
)

mn, mx = np.quantile(df.values, [0, 0.99])
d = mx - mn
ax1.set_ylim(mn-0.06*d, mx+0.15*d)
ax1.set_xmargin(0)

for name, series in df.items():
    sd = series.values.copy()
    s2 = series.sort_values()[95:]
    s1 = sd; s1[s2.index] = np.nan
    
    al, = ax1.plot(s1, label=name)
    ax1.plot(s2, '.', color=al.get_color())

    s1 = s1[~np.isnan(s1)]
    mu = np.mean(s1)
    sd = np.std(s1)
    std = sd / np.sqrt(sum(s1))

    print(*np.array([
        mu-2*std, mu, mu+2*std,
    ]).round(4))

    ax1.axhline(mu, ls='--', color='black', lw=1, alpha=0.5)
    kde = stats.gaussian_kde(s1)
    x = np.linspace(mu-6*sd, mu+6*sd, 201)

    ax2.plot(kde.evaluate(x), x, color=al.get_color())

ax1.set_xlabel('Timer run')
ax1.set_ylabel('time (seconds)')

ax1.legend()
ax2.get_xaxis().set_visible(False)

sns.despine(ax=ax1)
sns.despine(ax=ax2, bottom=True)

fig.tight_layout()

fig.savefig('idx_vs_walrus.png', dpi=160)

been a while, forgotten about this!

Can you explain how to use your code? What do you need for idx_vs_walrus.csv ?

Think this is what I'd do:

1. start by running the code in patched function.py to define all the things
2. run the timing code in my SE answer
3. get the data into a Pandas dataframe: df = pd.DataFrame(out, columns=[fn.__name__ for fn in functions])
4. save that to the CSV file: df.to_csv('idx_vs_walrus.csv', index=False)
5. run plot.py

note that at step 3 you'll have df loaded and don't need to save and reload it with pd.read_csv, but you can if you want to

Thank you !
Do you have any documentation that explain how to do what you do in patched walrus ? How it works, etc ?

Thank you !

no probs, happy to help

Do you have any documentation that explain how to do what you do in patched walrus ? How it works, etc ?

nothing apart from the documentation that's already on line. I'd suggest looking at Python's dis module to know what it's doing. i.e. try disassembling the functions in patched function.py to see what OP codes they get compiled down to

my code in patch was made up while answering the question. I've never tried to do anything like it before so there might well be better ways to go about doing it

Thank you :)