[question] can only be used with a log that provides a directory

[nichoth]

I was wondering if anyone could explain this restriction, where the flumelog must have a local directory. It seems like the dir name is only used to create the location for the leveldb, but is there another reason for this design?

  function create() {
      closed = false
      if(!log.filename)
        throw new Error('flumeview-level can only be used with a log that provides a directory')
      return Level(path.join(dir, name), {keyEncoding: charwise, valueEncoding: 'json'})
    }

https:/flumedb/flumeview-level/blob/master/index.js#L27

[dominictarr]

normally, you have a flumelog which is a file, and then a bunch of flumeviews, which also probably have files. To avoid a configuration nightmare, the views just created files or directories based on their name, next to the log. flumelog-memory doesn't have a log, it just creates a temporary log in memory. some other views (such as flumeview-reduce) notice this, and when used with a memory log, provide a memory view, but flumeview-level doesn't have an easy way to do this, so it throws an error instead, because it doesn't know where to put the file.

really the memory log thing is just for testing, but to make sure that it really really works, I usually test with a real log anyway.

[christianbundy]

Would it be appropriate to default to require('os').tmpdir() instead of throwing an error?

[dominictarr]

@christianbundy yes that would work. sometimes maybe print something in that case though because accidentially passing a memory log when you ment not to could be a mistake