| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
COUCHDB-1994
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
Similar change exists in the bigcouch branch, except we continue to use
a rebar.config file.
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
To set a release tag you can now pass a RELEASE_TAG parameter to the
make rel command (or via the env). The release tag will then be appended
to the version under the format <VERSION>-<TAG>.
If RELEASE_TAG=git then the latest short commit ID will be used and the
version will become <VERION>+build.<TAG>.
|
| |
|
| |
|
|
|
|
| |
user ERL_FLAGS instead of setting the path in the shebang
|
|
|
|
| |
tests on travis failed because deps aren't fetched.
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
tests were crashing in travis afte lager has been added: goldrush and
its dependencies were not started.
|
|
|
|
|
|
|
|
|
|
|
| |
For some reason noone noticed until now, but changing the log file was
only possible in the app.config until now. This change fix this by doing
the following:
- when couch_log starts it check if the log file handler has already
been started using the app.config. If not it start a new handler
- when the filename change after a config update using the API, the old
file handler is shutdown and a new one is started with the new config.
|
|
|
|
|
|
|
| |
Noticed when running tests vi travis-ci, jiffy doesn't like too much
lager and the build hangs. So instead to compile all modules with the
parse transform, only do it for couch_log. This is the only place where
we really need it anyway.
|
|
|
|
|
|
|
|
|
|
| |
With this change instead to use our own way to log file we are using
lager which improve the logging which will make couchdb more tolerant in
the face of large or many log messages, won't out of memory the node.
Note: Lager can handle multiple backend butwe are for now only handling
the file and console backend in the ini file. Other backends can be
configured using the app.config file.
|
|
|
|
|
| |
the record rep has bene upddated in the previous change to handle a
replication based on a view index. Fix it.
|
|
|
|
|
|
|
| |
fix build issue with some old version of rebar. Now the version of the
couch.app is set manually.
RCOUCH-2 done
|
|
|
|
|
|
| |
Rather than fetching the threshold and refresh interval settings each
time we need them, register the process to couch_config events so we can
update the config only when needed.
|
|
|
|
|
|
|
|
|
| |
Instead of a database, the replicator can now filter the documents using
a view index. All documents having a key emitted in the view can be
replicated.
View parameters can be used. Which means that you can replicate results
corresponding to a key in a view or a range.
|
| |
|
|
|
|
|
| |
Add couch_mrview:get_view_info/3 to retrieve the internal informations
of a view like the last update seq in this view or the number of rows.
|
|
|
|
|
| |
Make sure to only emit deleted document when a deleted key is passed to
the view filter.
|
|
|
|
| |
proper test to check the indexer.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This change add the possibility to trigger a view indexation in
background. The indexation can only work in background if at least one
process acquired it using the `couch_index_server:acquire_index/3`
function. If all the process that acquired it are down or released it
using `couch_index_server:release_indexer/3` then the background task is
stopped.
By default the background indexation will happen every 1s or when 200
docs has been saved in the database. These parameters can be changed
using the options `threshold` and `refresh_interval` in the couch_index
section.
To use it with couch_mrview a new option {refresh, true} has been added
to couch_mrview_changes:handle_changes Also the query parameter
refresh=true is passsed in t the HTTP changes API.
|
|
|
|
|
|
| |
If use_index=no even if the view is indexed by sequence, the index won't
be use. Instead it will fold the btree and return the changes each time
the view map function can emit a value. (default behaviour).
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Now when the option `seq_indexed=true` is set in the design doc, the
view filter in _changes will use it to retrieve the results. Compared to
the current way, using a view index will be faster to retrieve changes.
It also gives the possibility to filter changes by key or get changes in
a key range. All the view options can be used.
Note 1: if someone is trying to filter a changes with view options when
the views are not indexed by sequence, a 400 error will be returned.
Note 2: The changes will only be returned when the view is updated if
seq_indexed=true
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Similar to couch_changes:handle_changes but for view changes. It add
support for longpolling, normal and continuous stream The API differs
from the one for doc by beeing independant from the transport: the
support of HTTP will be added on top for example.
This API will be also used to replace the view filter in the current _changes
API.
Also add unittests.
|
| |
|
|
|
|
| |
See https://issues.apache.org/jira/browse/COUCHDB-1953
|
|
|
|
| |
COUCHDB-2031
|
|
|
|
|
|
|
| |
This refactor executes normalize_path/1 before appending the bound query
string parameters.
COUCHDB-2031
|
|
|
|
| |
function to refresh a view index.
|
| |
|
|
|
|
| |
test view changes and fix errors.
|
|
|
|
|
| |
This change smakes sure to compact the view indexes when using the view
changes.
|
|
|
|
|
| |
Add function to couch changes in a view index, similar to
couch_db:count_changes_since/2 .
|