dashron / roads-models Goto Github PK

add a "parallel" function, kinda like promises.

model1.load(id)
    .parallel(function () {
        return model2.load(id);
    })
    .ready(function (item1, item2) {

    })

I don't necessarily like the spec of he ready callback, code might be overly complex to achieve the correct order.

memcache
postgres

Predefine simple collections so that we can auto cache update them.

UserModule.collections['all'] = 'select id from users';

default onSave updates the all collection.

if (callback._error) {
callback._error(err);
}
else {
callback(err);
}

don't force users to use bind. it adds extra unnecessary computation. we can assume the _error pattern exists because we often will be chaining request promises

Sorts are applied after the onSave callback is fired, so if you end in save.ready() the sorts will never execute

Look into the mysql pooling and....

http://archive.coreservlets.com/Chapter18.html

http://publib.boulder.ibm.com/infocenter/iseries/v5r4/index.jsp?topic=%2Frzaha%2Fpoolnods.htm

on getConnection()....

if connection[0] is connected and ready, perform request
if connection[0] is connected and not ready, if connection.length <
max_connections, create connection and perform request, otherwise
queue up request
if connection[0] is not connected, shift off array and attempt the new
connection[0]

model.js: turns the sort into an order by statement

cachedmodel.js: uses an alternate cache key, which uses a sorted set
This sort stuff should be built into the buildCacheKey function so that onSave we can keep the sorted set up to date.
let mysql do the initial sorting for now.

so that it goes through the validation handler and doesn't end up being a db error
it would be cool if we could also bubble mysql errors through the validation handler

model.js : should simply turn the pagination into a limit

cachedmodel.js: should be filtered at the locateModels level. Keep every id cached in the set, but filter down the data when you pull the full model data.