so far we only had "StreetIndex" and "PoiIndex" only, and which to
use was determined by the chosen layout and whether a POI file was
attached to the job
in the future there will be multiple different indexers to choose
from, e.g StreetIndex and HealthIndex, and which one to use needs
to be maintained as part of the rendering job data
* maprenderingjob model now has a "queue" name column,
defaulting to "default"
* methods dealing with the rendering queue now all either
use the jobs queue column value if already available
or have an additional optional "queue_name" parameter
with default value "default"
* a systemd template service file was added to be able to start
individual per-queue render services
* the progress bar in the map request detail view was finally
fixed to show actual position-in-queue progress
these are specific to the local renderer configuration and
can differ a lot between instances, also git is not really
a good place to track binary files anyway
This is still far from perfect, but users should never see this anyway.
At least it now gives a hint towards what actually went wrong and not
just a very generic Django error message
HTML5 email vaildation is only triggered by an actual <input type="submit">
button being pressed, calling the forms submit() method from a regular
<button> does not. So we now have a hidden "real" submit button and
the actual styled button now sends a click even to that element instead
of submitting the form directly
PS: thinking again the actual issue may be that the styled button is not
inside the actual <form>?
so far when hitting the results directory threshold only few,
often just one, results were purged which basically triggered
another purge on the very next request rather often
as purge processing takes quite some time with large result
directories (somehting that needs solving, too, at a later time)
being near the threshold size could lead to substantial stalls
between individual render requests being processed
now purging will remove more results at a time, bringing the
directory size down to 90% of the threshold size whenever the
threshold is exceeded, and actual purge runs happen less often