In [1]:
import numpy as np
import pandas as pd
from utils import (
Adblocker,
compare_CDF,
barplot,
pieplot,
get_99p,
generate_table
)
In [2]:
brave = Adblocker(filename='data/brave_timings.json', color='#ff5500', label='Brave')
cliqz = Adblocker(filename='data/cliqz_timings.json', color='#00aff0', label='Cliqz')
ublock = Adblocker(filename='data/ublock_timings.json', color='#800000', label='uBlock Origin')
duck = Adblocker(filename='data/duckduckgo_timings.json', color='#58b792', label='DuckDuckGo')
abp = Adblocker(filename='data/adblockplus_timings.json', color='#c70d2c', label='Adblock Plus')
url = Adblocker(filename='data/url_timings.json', color='#1f3757', label='URL baseline')
tldts = Adblocker(filename='data/tldts_timings.json', color='#98b3be', label='Tldts baseline')
adblockfast = Adblocker(filename='data/adblockfast_timings.json', color='#f10c2b', label='AdblockFast')
Composition of Requests¶
In [3]:
pieplot(cliqz, ublock, duck, brave, abp)
Looking at all requests¶
Cliqz vs uBlock vs Adblock Plus vs Brave vs DuckDuckGo¶
In [4]:
compare_CDF(
cliqz,
ublock,
brave,
duck,
abp,
attr="all",
log_x=True,
log_y=False,
title='Decision Time: All Requests (CDF)'
)
Cliqz vs uBlock vs Adblock Plus vs Brave vs DuckDuckGo + Baselines¶
In [5]:
compare_CDF(
cliqz,
ublock,
brave,
duck,
abp,
tldts,
url,
attr="all",
log_x=True,
log_y=False,
title='Decision Time: All Requests (CDF)'
)
Cliqz vs AdblockFast¶
In [6]:
compare_CDF(
cliqz,
adblockfast,
attr="all",
log_x=True,
log_y=False,
title='Decision Time: All Requests (CDF)'
)
Interestingly, although AdblockFast is not a full-fledged blocker, and only uses 7 RegExp, Cliqz still perform as well.
In [7]:
compare_CDF(
cliqz,
ublock,
attr="all",
log_x=False,
log_y=False,
xlim=(0, 0.10),
title='Decision Time: All Requests (CDF)'
)
2. Cliqz vs AdblockPlus¶
In [8]:
compare_CDF(
cliqz,
abp,
attr="all",
log_x=True,
log_y=False,
title='Decision Time: All Requests (CDF)'
)
Comparing adblockers on requests that are NOT blocked¶
Reminder: This is the more common case: 80% of the requests are not blocked
In [9]:
compare_CDF(
cliqz,
ublock,
brave,
duck,
abp,
attr="not_blocked",
log_x=True,
log_y=False,
title='Decision Time: Not Blocked Requests (CDF)'
)
Comparing adblockers on requests that are blocked¶
In [10]:
compare_CDF(
cliqz,
ublock,
brave,
duck,
abp,
attr="blocked",
log_x=True,
log_y=False,
title='Decision Time: Blocked Requests (CDF)'
)
99th percentile¶
In [11]:
p99 = pd.DataFrame({
'request_type': ['_all', 'not_blocked', 'blocked'],
'cliqz': get_99p(cliqz),
'ublock': get_99p(ublock),
'brave': get_99p(brave),
'duck': get_99p(duck),
'abp': get_99p(abp),
})
p99['ublock_vs_cliqz'] = p99.ublock / p99.cliqz
p99['brave_vs_cliqz'] = p99.brave / p99.cliqz
p99['duck_vs_cliqz'] = p99.duck / p99.cliqz
p99['abp_vs_cliqz'] = p99.abp / p99.cliqz
In [12]:
p99
Out[12]:
| request_type | cliqz | ublock | brave | duck | abp | ublock_vs_cliqz | brave_vs_cliqz | duck_vs_cliqz | abp_vs_cliqz | |
|---|---|---|---|---|---|---|---|---|---|---|
| 0 | _all | 0.047663 | 0.035182 | 0.027854 | 12.594179 | 0.076995 | 0.738141 | 0.584395 | 264.233871 | 1.615404 |
| 1 | not_blocked | 0.043227 | 0.034014 | 0.031603 | 11.960517 | 0.077617 | 0.786869 | 0.731094 | 276.690888 | 1.795568 |
| 2 | blocked | 0.050730 | 0.036822 | 0.023256 | 13.131594 | 0.076321 | 0.725843 | 0.458427 | 258.852632 | 1.504455 |
Medians¶
In [13]:
medians = pd.DataFrame({
'request_type': ['_all', 'not_blocked', 'blocked'],
'cliqz': list(map(lambda x: np.median(x), [cliqz.all.raw, cliqz.not_blocked.raw, cliqz.blocked.raw])),
'ublock': list(map(lambda x: np.median(x), [ublock.all.raw, ublock.not_blocked.raw, ublock.blocked.raw])),
'brave': list(map(lambda x: np.median(x), [brave.all.raw, brave.not_blocked.raw, brave.blocked.raw])),
'duck': list(map(lambda x: np.median(x), [duck.all.raw, duck.not_blocked.raw, duck.blocked.raw])),
'abp': list(map(lambda x: np.median(x), [abp.all.raw, abp.not_blocked.raw, abp.blocked.raw]))
})
In [14]:
medians['ublock_vs_cliqz'] = medians.ublock / medians.cliqz
medians['brave_vs_cliqz'] = medians.brave / medians.cliqz
medians['duck_vs_cliqz'] = medians.duck / medians.cliqz
medians['abp_vs_cliqz'] = medians.abp / medians.cliqz
In [15]:
medians
Out[15]:
| request_type | cliqz | ublock | brave | duck | abp | ublock_vs_cliqz | brave_vs_cliqz | duck_vs_cliqz | abp_vs_cliqz | |
|---|---|---|---|---|---|---|---|---|---|---|
| 0 | _all | 0.007496 | 0.010978 | 0.006168 | 8.044320 | 0.011308 | 1.464514 | 0.822839 | 1073.148346 | 1.508538 |
| 1 | not_blocked | 0.007620 | 0.011009 | 0.006223 | 2.164159 | 0.011163 | 1.444751 | 0.816667 | 284.010367 | 1.464961 |
| 2 | blocked | 0.007197 | 0.010934 | 0.006062 | 8.311366 | 0.011610 | 1.519244 | 0.842295 | 1154.837571 | 1.613172 |
In [16]:
for request_type in ['_all', 'not_blocked', 'blocked']:
print('Request type', request_type)
print(generate_table(p99, medians, request_type=request_type))
Request type _all
| | 99% OF REQUESTS | MEDIAN |
| ------------- | ---------------------------- | ---------------------------- |
| **Cliqz** | **0.048ms** | **0.007ms** |
| uBlock Origin | 0.035ms (**0.7x slower**) | 0.011ms (**1.5x slower**) |
| Adblock Plus | 0.077ms (**1.6x slower**) | 0.011ms (**1.5x slower**) |
| Brave | 0.028ms (**0.6x slower**) | 0.006ms (**0.8x slower**) |
| DuckDuckGo | 12.594ms (**264.2x slower**) | 8.044ms (**1073.1x slower**) |
Request type not_blocked
| | 99% OF REQUESTS | MEDIAN |
| ------------- | ---------------------------- | ---------------------------- |
| **Cliqz** | **0.043ms** | **0.008ms** |
| uBlock Origin | 0.034ms (**0.8x slower**) | 0.011ms (**1.4x slower**) |
| Adblock Plus | 0.078ms (**1.8x slower**) | 0.011ms (**1.5x slower**) |
| Brave | 0.032ms (**0.7x slower**) | 0.006ms (**0.8x slower**) |
| DuckDuckGo | 11.961ms (**276.7x slower**) | 2.164ms (**284.0x slower**) |
Request type blocked
| | 99% OF REQUESTS | MEDIAN |
| ------------- | ---------------------------- | ---------------------------- |
| **Cliqz** | **0.051ms** | **0.007ms** |
| uBlock Origin | 0.037ms (**0.7x slower**) | 0.011ms (**1.5x slower**) |
| Adblock Plus | 0.076ms (**1.5x slower**) | 0.012ms (**1.6x slower**) |
| Brave | 0.023ms (**0.5x slower**) | 0.006ms (**0.8x slower**) |
| DuckDuckGo | 13.132ms (**258.9x slower**) | 8.311ms (**1154.8x slower**) |
Serialization and Deserialization timings¶
In [17]:
compare_CDF(
cliqz,
ublock,
brave,
abp,
adblockfast,
attr="serializationTimings",
log_x=True,
log_y=False,
y_label="Fraction of All Samples",
title='Serialization Timings (CDF)'
)
In [18]:
compare_CDF(
cliqz,
ublock,
brave,
abp,
adblockfast,
attr="deserializationTimings",
log_x=True,
log_y=False,
y_label="Fraction of All Samples",
title='Deserialization Timings (CDF)'
)
In [19]:
def compute_loading_medians(field, *adblockers):
return pd.DataFrame({
'adblocker': [adblocker.label for adblocker in adblockers],
field: [np.median(getattr(adblocker, field).raw) for adblocker in adblockers],
})
In [20]:
serialization_medians = compute_loading_medians('serializationTimings', cliqz, abp, brave, ublock)
barplot(
serialization_medians,
'serializationTimings',
title='Serialization Timings (Median)',
ylabel="Time (ms)",
color=[cliqz.color, abp.color, brave.color, ublock.color],
save=True
)
In [21]:
deserialization_medians = compute_loading_medians('deserializationTimings', cliqz, brave, ublock, abp)
barplot(
deserialization_medians,
'deserializationTimings',
ylabel="Time (ms)",
title='Deserialization Timings (Median)',
color=[cliqz.color, brave.color, ublock.color, abp.color],
save=True
)
In [22]:
deserialization_medians
Out[22]:
| adblocker | deserializationTimings | |
|---|---|---|
| 0 | Cliqz | 0.359295 |
| 1 | Brave | 271.456332 |
| 2 | uBlock Origin | 17.000704 |
| 3 | Adblock Plus | 144.437919 |
Serialized Size¶
In [23]:
size_medians = pd.DataFrame({
'adblocker': [ublock.label, cliqz.label, abp.label, brave.label],
'Cache Size': [ublock.cacheSize, cliqz.cacheSize, abp.cacheSize, brave.cacheSize]
})
size_medians['Cache Size'] /= (10**6)
In [24]:
barplot(
size_medians,
'Cache Size',
ylabel="Size (MB)",
color=[ublock.color, cliqz.color, abp.color, brave.color],
save=True
)
Memory Usage¶
In [25]:
memory_medians = pd.DataFrame({
'adblocker': [adblockfast.label, cliqz.label, ublock.label, abp.label, duck.label],
'Memory Usage at Startup': [0.0013, 1.745, 2.525, 9.838, 15.911],
})
barplot(
memory_medians,
'Memory Usage at Startup',
ylabel="Size (MB)",
color=[adblockfast.color, cliqz.color, ublock.color, abp.color, duck.color],
save=True
)
In [26]:
memory_medians = pd.DataFrame({
'adblocker': [ublock.label, cliqz.label, abp.label, duck.label],
'Memory Usage after matching': [4.450, 5.575, 13.407, 15.962]
})
barplot(
memory_medians,
'Memory Usage after matching',
ylabel="Size (MB)",
color=[ublock.color, cliqz.color, abp.color, duck.color],
save=True
)
Time to Parse Lists¶
In [27]:
time2parse_medians = pd.DataFrame({
'adblocker': [abp.label, cliqz.label, ublock.label, duck.label, brave.label],
'Time to parse Easylist All': [abp.parsingTime, cliqz.parsingTime, ublock.parsingTime, duck.parsingTime, brave.parsingTime]
})
barplot(
time2parse_medians,
'Time to parse Easylist All',
ylabel="Time (ms)",
color=[abp.color, ublock.color, cliqz.color, duck.color, brave.color],
save=True
)
In [28]:
time2parse_medians = pd.DataFrame({
'adblocker': [abp.label, cliqz.label, ublock.label, duck.label],
'Time to parse Easylist Without Brave': [abp.parsingTime, cliqz.parsingTime, ublock.parsingTime, duck.parsingTime]
})
barplot(
time2parse_medians,
'Time to parse Easylist Without Brave',
ylabel="Time (ms)",
color=[abp.color, cliqz.color, ublock.color, duck.color],
save=True
)
