Noissue - Reimplement public suffix matching more efficiently

Noissue - Reimplement public suffix matching more efficiently

[Sebastian Noack, 2017-05-01 18:53:29]

What is the reason for this change? I'm not aware that we are bothered whether
this script runs for 1 or 10 seconds on the server, while this change makes the
code arguably more complex, hence less maintainable. Furthermore, we already
have this logic implemented in adblockpluschrome, so if at all, we should rather
reuse that implementation before rolling yet another one. But again, I'm missing
a case, that justifies an optimization here.

[Manish Jethani, 2017-05-01 19:00:50]

On 2017/05/01 18:53:29, Sebastian Noack wrote:
> What is the reason for this change?

While working on issue #3673, I discovered that the script was spending most of
its time in the tlds.getDomain function. I looked into tldjs: it is terribly
inefficient.

Now with this change, I have reimplemented base domain extraction using the
Public Suffix List.

This speeds up the script by ~85% for easylist.txt.

> I'm not aware that we are bothered whether
> this script runs for 1 or 10 seconds on the server, while this change makes
the
> code arguably more complex, hence less maintainable.

I got the impression from the comments on issue #3673 that performance was
important for the generation of the iOS filters. For example, my implementation
for #3673 nearly doubles the time it tasks to generate the filters. If that is a
concern, this optimization more than makes up for that extra time spent.

[Sebastian Noack, 2017-05-01 19:10:24]

On 2017/05/01 19:00:50, Manish Jethani wrote:
> I got the impression from the comments on issue #3673 that performance was
> important for the generation of the iOS filters.

While #3673 is about an optimization, what we are optimizing for there is the
size of the list for the client-side. I don't see how this is related to the
time on the server-side spend in order to generate the list, or do I miss
something?

[Sebastian Noack, 2017-05-01 19:14:42]

On 2017/05/01 19:10:24, Sebastian Noack wrote:
> On 2017/05/01 19:00:50, Manish Jethani wrote:
> > I got the impression from the comments on issue #3673 that performance was
> > important for the generation of the iOS filters.
> 
> While #3673 is about an optimization, what we are optimizing for there is the
> size of the list for the client-side. I don't see how this is related to the
> time on the server-side spend in order to generate the list, or do I miss
> something?

What, however somewhat matters, and I think this is what you are referring to,
is the performance when this code is used on the client-side in Adblock Plus for
Safari on macOS. However, in this case we don't use the tldjs module for node,
but our own implementation (from adblockpluschrome) which is already
significantly faster.

[Manish Jethani, 2017-05-01 19:50:19]

On 2017/05/01 19:14:42, Sebastian Noack wrote:

[snip]
> is the performance when this code is used on the client-side in Adblock Plus
for
> Safari on macOS. However, in this case we don't use the tldjs module for node,
> but our own implementation (from adblockpluschrome) which is already
> significantly faster.

That's interesting, I'll look into it.

[Manish Jethani, 2017-05-11 14:04:26]

Closed.