Mirror of forum.nim-lang.org

2641 :: InfluxDB-to-MySQL Protocol Converter

• pwernersbach (orginal) [2016-11-30T07:18:53+01:00] view original

  I've open sourced and published my InfluxDB-to-MySQL protocol converter to https://github.com/philip-wernersbach/influx-mysql. It is written in Nim, and is being used in production to power a big data solution. This is a great example of production software written in Nim.

---

• gokr (orginal) [2016-11-30T18:13:16+01:00] view original

  Yikes, that's a lot of code! Impressive. And yes, a good solid example.

---

• pwernersbach (orginal) [2016-12-01T05:44:50+01:00] view original

  Haha, yep, it's a pretty comprehensive project and is bigger than a lot of Nim projects. The code makes use of Nim features such as async and macros. Imagine how large the program would be if Nim didn't have those features!

---

• perturbation2 (orginal) [2016-12-04T04:48:37+01:00] view original

  @pwernersbach - That's a really cool project, and I'm glad that you shared it! I love that you use the 'not nil' annotations for your argument types and type definitions, and I've enjoyed reading through it, even though I'm not currently using InfluxDB.

  I hope you don't mind a few questions, but I'd really like to learn more about how Nim is used "in the wild" since I'm using it for my own projects and would love to learn more best practices.

  • Could you talk a bit about the useDB macro? I can see that it's a try-with-resources for the DB connection, but could this be done as a template (like the 'withFile' example in the manual?) Is using the immediate pragma needed so that you don't have to mark all of the parameters as 'untyped', or are you doing that for the overloading resolution?
  • How did using nimprof work for you? Is that why / how you have the GC_disable and GC_enable in critical parts?
  • The problem you point to with refcounting and long link lists seems bad and I definitely see the rationale for reflists - it looks like you're using finalizers with ReadLinesFutureContext and its 'super' to manage cleanup - have you had any problems with finalizers? I'm not sure I've seen another project using Nim finalizers in production, and the manual talks about using destructors differently, so I'm not sure if the new(foo, destructor) is here to stay or not.

  • In your splitIndividualStatements iterator, you're checking for 'SELECT' character-by-character - could you do something like

    if stmts[pos + 1 .. pos + 6] == "SELECT":
         # Do whatever work here
         discard
       
    Run

    or was there a performance difference in doing character-at-a-time approach?

---