Today, I have released DaCHS 1.1, with the main selling point that DaCHS should now speak TAP 1.1 (as defined in the current draft).
First off, if you're not yet on DaCHS 1.0, please read the corresponding release article before upgrading.
As usual, the general upgrading instructions are available in the operator's guide (in short: do a dachs val ALL before the Debian upgrade). This time, I'd recommend to use the opportunity to upgrade your underlying server to stretch if you haven't done so already. If you do that, please have a look at hints on postgres upgrades. Stretch comes with postgres 9.6 (jessie: 9.4). Postgres upgrades are generally safe, but please take a dump before migrating anyway.
So, with this out of the way, here's a short list of the major changes from DaCHS 1.0 to DaCHS 1.1:
- DaCHS now officially requires python 2.7. If this really is a problem for you, please shout – if wouldn't be hard to maintain 2.6 compatibility, but by now we feel there's no reason to bother any more.
- Now supporting TAP 1.1; in particular, TOP n doesn't trump MAXREC any more, and it doesn't affect OVERFLOW indication, which may break things that used TOP to override DaCHS' default TAP match limit of 2000. Also, TAP_SCHEMA is updated (this happens as a side effect of dachs upgrade).
- Now serialising spoint, scircle, and friends to DALI 1.1 xtypes (timestamp, point, polygon, circle). Fields explicitly marked with adql:POINT or adql:REGION will still be serialised to STC-S. Do this only if you have no choice (DaCHS has this for obscore and epntap s_region right now).
- The output column selection is sanitised. This may make for slight changes in service responses, in particular in VOTable formats. See Output Tables in the reference documentation for details if you think this might hit you.
- DaCHS no longer comes with an outdated version pyparsing and instead uses what's installed on the system. The Debian package further re-uses additional system resources if available (rjsmin, jquery).
- DaCHS now tries a bit harder to come up with sensible names for SODA result files.
- map/@source is no longer limited to identifier-like strings; any key that's in your source is fair game.
- For incremental imports with data that's updated now and then, there's now ignoreSources/@fromdbUpdating.
- Relative imports from custom code ("import foo" in a custom core, for instance, getting res/foo.py) no longer work. See Importing Modules in the reference documentation for details.
- This release fixes a severe bug in the creation of obscore metadata from SSAP tables. If you use //obscore#publishSSAPHCD or //obscore#publishSSAPMIXC mixins, update the obscore definitions by running dachs imp -m <rdid>, followed by dachs imp //obscore (the latter is only necessary once at the end).
- You can now define a footer.html template that's added at the foot of the main page content – with a bit of CSS magic, this lets you overwrite almost anything on DaCHS HTML pages.
As always, please complain early if something breaks for you; our regression tests can only cover so much. In particular, our support list is there for you.
Update (2017-12-06): In particular on jessie, you may see that all DaCHS packages are being held back. To resolve this situation, manually say apt-get install python-gavoutils python-gavostc.