Re: ordered string-list considered harmful, was Re: [PATCH v3] Allow aliases that include other aliases

From: Jonathan Nieder <jrnieder@gmail.com>
To: "Ævar Arnfjörð Bjarmason" <avarab@gmail.com>
Cc: Jeff King <peff@peff.net>,
	git@vger.kernel.org, Tim Schumacher <timschumi@gmx.de>,
	gitster@pobox.com, pclouds@gmail.com
Subject: Re: ordered string-list considered harmful, was Re: [PATCH v3] Allow aliases that include other aliases
Date: Fri, 7 Sep 2018 00:23:53 -0700	[thread overview]
Message-ID: <20180907072353.GC172953@aiede.svl.corp.google.com> (raw)
In-Reply-To: <87in3hrcya.fsf@evledraar.gmail.com>

Ævar Arnfjörð Bjarmason wrote:
> On Fri, Sep 07 2018, Jonathan Nieder wrote:
>> Jeff King wrote:

>>> I don't see any point in generating a sorted list and _then_ making an
>>> auxiliary hashmap. My idea was that if you're using a sorted string-list
>>> for lookup, then you can replace the whole thing with a hash (inserting
>>> as you go, rather than sorting at the end).
>>
>> What if I'm sorting a string list in preparation for emitting a sorted
>> list, and I *also* want to perform lookups in that same list?  In
>> other words:
>
> If this turns out to be a common use-case perhaps the easiest way to
> support that would be to make the hashmap (optionally?) ordered, as Ruby
> 1.9 did with their hash implementation:
> https://www.igvita.com/2009/02/04/ruby-19-internals-ordered-hash/

That's about recording the order of insertion.  I'm talking about
something much simpler: sorting an array (as preparation for emitting
it) and binary searching to find an entry in that array.