Overpass API developpement

[mmd <>]

Hi,

Am 27.02.2015 um 20:47 schrieb Donal Diamond:

> Has the new  --flush-size=0  argument something to do with this?

the recently added --flush-size=0 seems to set the flush limit to a very
high value, so that pretty much all of the changes in a minutely diff
will be collected in main memory, before they are eventually flushed to
disk.

I would assume that this is an optimization to avoid repeated
compression/decompression steps during updating - at the expense of
additional main memory consumption. But of course, that alone doesn't
really explain larger file sizes.

BTW: Did you already run some queries on your compressed db? Is there a
noticeable performance impact compared to the non-compressed db?

Best,
mmd