How this does work?


This is about fixed length data time series. For the page about time series, see the proper page.

Data is stored in so called chunks. A chunk’s last page can be actively appended to, or a chunk is immutable.

When there is a request to fetch some data, a chunk is loaded into memory. It will not be automatically unloaded, to do this, you must periodically call close_chunks().


Start off by instantiating an object

Note that if you specify a gzip_level argument in create_series(), GZIP compression will be used.


Note that gzip-compressed series are very slow to read, since every seek is conducted from the beginning. indexed-gzip library hung too often. This will be fixed in the future.

Also, any gzip-opened series will raise a warning, since their support is experimental at best.

You can create new databases via

Then you can create and retrieve particular series:

You retrieve their data via Iterators:

Appending the data is done via append(). Since time series are allocated in entire pages, so your files will be padded to a page in size. This makes writes quite fast, as in 99.9% cases it is just a memory operation.


tempsdb will log when opening and closing series. To prevent this from happening, just call: