cpuminer-opt API

[JayDDee]

I have avoided messing with the API because I'm not a web programmer. The enhanced statistics now reported to the console would be very useful if accesible through the API.

There are several miners with APIs and I;m sure some are better than others. There may also be some quasi standards for miner APIs.

This issue is opened for suggestions on improvement. Examples of the best existing implementations, preferably with open source code would be great.

Ideas on how the statistics should be provided are also welcome. The ability to stream the data rather that the existing one shot command would be nice, but I would probably need some technical help with that.

One addition I intend to make is version control to facilitate changes without forcing users to upgrade their scripts promptly.

The floor is open.

[JayDDee]

One if the big concerns about adding more features is the impact on mining performance. Adding a streaming feature to the API would likely have an impact.

Instead it might be preferable to use a pipe to redirect miner output to an external program that could parse the data for analysis or transmission to a remote manager.

The key would be a stable output format with certain requirements to aid parsing while maintaining human readability. This would avoid any additional overhead in the miner.

[YetAnotherRussian]

Up to 90% of users do disable the API in xmrig/xmrstack (some forks even generate config files with disabled api by default). The rest (powerful miners) uses HiveOS or similar (so there is no interest in miner api as well). My guess is that "avoided messing with the API" was the right way of doing things there.

Not bad API is here: https://github.com/Bendr0id/xmrigCC/tree/master/src/base/api (though some issues have been fixed in the parent repo).

[offtopic] Sorry for being aware of #246, I was out for a week, but now I cannot sync, the test coin seems to be down, even resync from scratch does not help, or maybe I got banned @ nodes (I still see the pool being alive though). I'll search for another low-hr victim somewhen between this and then. There're some I currently hang on, but you've dropped cryptonight support (so I see no reason in testing with a veeery old releases).

[JayDDee]

Using the API as as seperate data stream is inefficient. A better approach might be to use a pipe to redirect stdout to another program that can parse the data and process it as it pleases.

If the current data is machine readable as well as human readable in its curent form then no changes would be required to the miner and all overhead would be borne by other programs.

[JayDDee]

I'm closing this. There doesn't seem to be much interest and I don't have a clear vision or direction. Any attempt at a streaming feature would add additional overhead that would affect performance.