Cache address from ScriptPubKey

[ianadavies]

Performance Improvement. - This PR caches the address read from the ScriptPubKey when using the ScriptAddressReader. This is currently done for every HDPayment Record and is an expensive operation.

[codingupastorm]

Tested this. It seemed to be slower with manual profiling (stopwatching, average of 5 queries don't judge lol).

My GetHistory call with 10k entries and ~215 sends went from ~2.8s to ~3.5s.

I looked into this and tried to optimise... Avoid the ToLists and multiple settting on the dictionary, removed the lookup entirely and made the cache a normal Dictionary, we don't need concurrent.

New code would look like this:

private static readonly Dictionary<Script, string> ScriptPubKeyCache = new Dictionary<Script, string>();

            if (transactionData.SpendingDetails == null || this.ScriptAddressReader == null)
                return res;

            foreach (var paymentDetails in res)
            {
                if (ScriptPubKeyCache.TryGetValue(paymentDetails.DestinationScriptPubKey, out string storedAddress))
                {
                    paymentDetails.DestinationAddress = storedAddress;
                }
                else
                {
                    string address = this.ScriptAddressReader.GetAddressFromScriptPubKey(this.Network, paymentDetails.DestinationScriptPubKey);
                    ScriptPubKeyCache[paymentDetails.DestinationScriptPubKey] = address;
                    paymentDetails.DestinationAddress = address;
                }
            }

            return res;

Using the above code, the GetHistory call was marginally faster. Seemed to be ~2.6s.

[codingupastorm]

But I don't think those results are significant enough to warrant putting in at this stage.

[zeptin]

I concur with @quantumagi, if there is a worthwhile performance benefit from introducing this cache, then I would suggest utilising an LRU cache implementation with a bounded maximum size. See the MemorySizeCache / MemoryCountCache in Stratis.Bitcoin.Utilities as possible ready-made implementations.

[codingupastorm]

3.0.6.0 is closed, so I think we can close this. If you're still keen to get this in @ianadavies feel free to open a PR to 3.0.7.0