<div class="gmail_quote">On Fri, Nov 19, 2010 at 12:21 AM, Andreas Kalsch <span dir="ltr"><<a href="mailto:andreaskalsch@gmx.de">andreaskalsch@gmx.de</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">
It's always great to try something new, but we use Postgres as the database and it is SQL, where I use tables. Somehow I feel forced to learn something I and others will not use very often. It is better to be conservative about schemas. The problem: I have written some software that heavily relies on the 0.36 simple schema. Now I have to use 0.36 and cannot use the new PBF format.<br>
It would be good to be able to choose between hstore and the old schema that did my job very well.<br>
What about GROUP BY over single hstore k/v pairs - is this possible? hstore feels like having a table inside a cell ...<br></blockquote><div><br>If you don't need the performance or features provided by hstore, then stick with Osmosis 0.36 for now. If you do need new features like PBF binary support you can run two osmosis installs side-by-side and pipe between them if necessary. However if you decide to process a full planet I think you will appreciate the performance gains that hstore provides.<br>
<br>Can you give an example of a GROUP BY query you were performing before? Perhaps we can figure out an alternative using the new hstore data type.<br><br>Note that while the schema is compatible with PostgreSQL 8.3 and later, newer versions provide additional functions that make it easier to convert from hstore to sets and vice versa.<br>
<br>Brett<br></div></div><br>