评估 10,000 条规则的最快算法是什么？

Question

I have pricing rules to determine how much discount I would give to customer. Problem is I need lots of rules, like ~10,000 rules, and the performance would be very slow if I were to loop all 10,000 rules for each customer request.

There are many conditions that I need to check to apply certain discount:

- Product type (clothes, electronics, etc)
- Product SKU
- Customer location
- Search date (e.g. >= 2019-01-01 And <= 2019-01-31)
- .
- .
- .
- ~30th conditions

Here are the example of rules I'd like to set:

Rule 1: product type = 'clothes', then discount 10%
Rule 2: product type = 'electronics', then discount 5%
Rule 3: product type = 'clothes' AND customer location = 'AUSTRALIA', then discount 7%
.
.
.
Rule 10,000: ....

Also I want to make each rule have priority. So if Rule 3 has higher priority than Rule 1, I want to apply discount using Rule 3.

The naive approach would be to loop all 10,000 rules and check each of the rules one by one whether they match the conditions or not. But the performance would be very bad. And what if I want to add another 10,000 rules.

I'm very interested to know if there is better approach for this instead of looping all rules.

--- update This rule will need to be triggered everytime a user do search. There's a search bar which user can type the keywords he wants to find, and the page will return all the products that match the keywords. The result could reach up to 50 products, hence 50x we need to evaluate which rules applies to each product for each user search.

Answer 1

This may be a bit of an overkill, but when I think speed, I think Hash Tables , where unique rules are stored as (rule, discount) pairs.

For this to work, you'll need to categorize your rule criteria (product type, country, etc). Second, you'll need to assign a number (enumerate) each member of each category:

Countries[Australia = 1, New Zealand = 2, ...]

After that, split all of the rules with multiple acceptable criteria into separate rules:

Rule 3: product type = 'clothes' AND (customer location = 'AUSTRALIA' OR customer location = 'NEW ZEALAND'), then discount 7%

becomes

Rule 4: product type = 'clothes' AND customer location = 'AUSTRALIA', then discount 7%
Rule 5: product type = 'clothes' AND customer location = 'NEW ZEALAND', then discount 7%

Now you have an array of criteria to check. If none are specified, you can leave a zero. For instance for array of criteria:

[product type, customer location, month]

you can have values

['decorations', '', 'December']

which translate to

[23, 0, 12]

and if you have a total of, say, 8 types of criteria you want to check, you have a final array that looks like this

[0, 0, 0, 23, 0, 0, 12, 0]

Now is the time to check which is the specific rule that applies to this, by performing the hash function H() on the array in some form. You could just string the digits together:

=H(0002300120)

or you could multiply each successive number with a greater power of 10 and then add them together (for less than ~25 criteria, because of 2 ⁶⁴ limit):

=H(230000 + 120000000)

The beauty of Hash Tables is that they work almost instantly with O(1) , if they are made with enough space to begin with and have both good hash functions H() and collision resolving mechanism (because not all H() may be unique each time).