Matthieu Muffato authored 11 years ago and Andy Yates committed 11 years ago

Compara v74: the taxon_id and taxon_name tags are deprecated, we should use the species_tree_node() call instead

We had an error whenever we used the ensembl genomes species with the
multi databases from compara. Our best guess compara DB was switching
to the division’s compara DB rather than using multi. We now avoid
that problem by giving a default. So we will still try to use the
division’s DB but the lack of it will result in the correct results
being passed back.

* taxonomyEndpoint:
  Taxonomy name endpoint created, basic tests for all taxonomy endpoints.
  Preservation commit for taxonomy name endpoint

we can not return anything
this can be because the id never existed in ensembl, then say that it could not be found in lookup
but it can also be because no lookup database was found and other means did not return any result
in that case, inform the user of the absence of the database
also, when using versioned stable id, use stable id without version for quick lookup, then version for archive retrieval

archive test module now uses lookup test database for archive lookup

Before we had some bizarre behaviour by trying to query for a MLSS with the same species mentioned twice. That
no longer happens. We can query for self homologies if needed. Also we do not try to find bad MLSS sets.

We fixed a single : error when detecting OntologyXref objects. We also added some IdentityXref tests just to avoid any future regressions

Seems that we had an issue highlighted by Matthias Ziehm about specifying more than one target taxon. This was totally wrong and a bug in the REST API. This fixes the bug

The previous solution was non-optimal and missed out YAML and Sereal formats. This
does not.

We now look at the type of method and will switch to looking at the Accept header.

Finally we have a JSON output for gene trees. it's not done nicely but it works. Also it is very very fast! Support is done using the TreeHash builder and then delegating to the JSON encoder. A custom serialiser has been built in the view layer. Otherwise we would have to do content-type/accepts detection. That's not nice at all!

We had the minute test fail I'm guessing because we elapsed through the minute. So instead we sleep until the new minute. It's a dodge but better than nothing

We used to extend existing code to provide decent throttling. Lets DIY it and we can have a far nicer way of doing it.

We needed a better rate limiter for our system. This is more suited to Ensembl's needs and means we can customise the headers returned (new work suggests even though HAProxy/Nginx can provide limits on the number of requests we cannot do very complex limits).

We now use the preload() function on a GeneTree to fetch lots of info
from the server. Local tests suggest it wins back half a second or so. If anything
it is fewer trips to the server

We have reconstructed the exon splicing in BED and enabled it for all feature endpoints.

We have the context() method which provides the current catalyst
context on each request. So there is little reason to pass $c
between methods (apart from the roles which need it still)

Only allowing this on a number of content types but we do allow it now.

endpoint uses object_type to restrict the query, not object