Independent writes: Make workaround more flexible

[franzpoeschel]

Follow-up to #1619
This lets users decide to flush only the containing Iteration or everything with Attributable::seriesFlush().

TODO:

• Maybe use false as a default even. Would be a breaking change, but I'd argue it's the more intuitive behavior and most usage is probably along that line. They're two different functions now.

[guj]

Thanks Franz.

1. It seems to me that for an I/O library that requires collective efforts, e.g. BP5, it is better to use iteration->seriesFlush() instead of series->flush(). So, should the default behavior of series->flush() be flushing everything, and set a flag like "dirtyonly= true" if optimization is desired.

2. naming is a bit confusing to me. by calling iteration->seriesFlush(flush_entire_series=true), it will flush all other iterations in the underlying series owning this iteration? might as well have something like series->flush(dirtyonly=false)? i.e. I think letting iteration/series flush itself only would be cleaner.

[ax3l]

Good points. I agree, we could make it maybe more transparent by introducing iteration->flush() and keeping iteration->seriesFlush() to point to the series flush?

[franzpoeschel]

In that case, I'd suggest to add Attributable::iterationFlush() instead of Iteration::flush():

• more intuitively tells the meaning of Attributable::seriesFlush() in opposition to Attributable::iterationFlush()
• users don't need to carry around the Iteration for flushing