Comparing caching strategies

[avi.e.gross at gmail.com]

MRAB,

I made it very clear I was using the translation provided by Google Translate. I copied exactly what it said and as I speak the languages involved, they seemed reasonable.  I often find it provides somewhat different translations than I expect and sometimes I need to supply a longer sentence to get it on track and sometimes it is just plain wrong. Getting it to provide female-specific sentences can be an exercise in frustration, for example.

What it produces is not important to the main point. 

It is hard to have conversations when people keep niggling on details and especially details that make absolutely no difference. 

My point was you can have functions that cannot be cached for results just any trivial way. The EXAMPLE I gave suggests if your had a Python program that did something long these lines between multiple languages, the RESULTS will depend on not just the phrase used. But they can be cached well in several ways if you want.

Let me point out another related use. When you type a message on your phone, you may have a sort of prediction algorithm running that keeps offering to auto-complete your words or fix spelling errors. It does this dynamically and learns new words and eventually may start remembering patterns. I have totally mucked up my ENGLISH keyboard because it now remembers snippets of my typing in languages like Spanish and German and lately Esperanto and it makes suggestions from a very confused composite state including lots of funny accented characters. I now switch keyboards when I remember but much of the damage is done! LOL!

That too is an example of some hidden data the program uses which is loaded with past data of what I have typed and of short phrases so it can make a guess that word A is often followed by word B and after seeing both A and B is extremely likely to be followed by C. Now obviously this would not be a trivial use of caching as part of guessing but could be part of a method that increments some keys like "A B" or "A B C" with a count of how often they have seen that combo. Yes, this is not direct caching, but also a side effect you might want to add to a function.

As for Esperanto, I am still learning it and as a created language, the snippet I used can likely be translated multiple ways and as someone who could care less about Valentines Day, I don't ever intend on using any of those ways in conversation. Most languages have alternate ways of saying things and it would not shock me if there are sentences like "The Holiday used in Capitalist Western Countries to commemorate a day of love based on a person some religions consider of merit and have given a religious title" or whatever.

So back on topic, an original question here was how to cache things, perhaps using a LRU algorithm with a data structure using some maximum.

My comment was that using a function decorator that caches any result may not be adequate in many cases. I presented several examples and my point is that in the above example, it may make more sense to have multiple caches that exist perhaps outside any one function, or a more complex cache that stores using a more complex key

-----Original Message-----
From: Python-list <python-list-bounces+avi.e.gross=gmail.com at python.org> On Behalf Of MRAB
Sent: Saturday, February 18, 2023 7:04 PM
To: python-list at python.org
Subject: Re: Comparing caching strategies

On 2023-02-18 23:04, avi.e.gross at gmail.com wrote:
[snip]
> 
> Note how this can cause problems with the original idea here of caching strategies. Imagine a function that checks the environment as to what encoding or human language and so on to produce text in. If you cache it so it produces results that are stored in something like a dictionary with a key, and later the user changes the environment as it continues running, the cache may now contain invalid results. You might need to keep track of the environment and empty the cache if things change, or start a new cache and switch to it.  An example would be the way I use Google Translate. I sometimes am studying or using a language and want to look up a word or phrase or entire sentence. If Google Translate keeps track, it may notice repeated requests like "Valentines Day" and cache it for re-use. But I often click to switch languages and see if the other one uses a similar or different way to describe what it means or something similar but spelled another way. German does the latter as in Valentinstag which is a fairly literal translation as does Dutch (Valentijnsdag ) and  Hungarian (Valentin nap) .
> 
> But Hebrew calls it the holiday of love, sort of (חג האהבה). 
> Portuguese is similar but includes day as well as love (Dia dos 
> Namorados)
> 
> Esperanto tosses in more about sainthood (Sankta Valentín) and in a sense Spanish does both ways with day and saint (Día de San Valentín).
> 
The Esperanto didn't look right to me; it's "Valentena tago" or "Sankt-Valentena tago".

[snip]

--
https://mail.python.org/mailman/listinfo/python-list