MQuery-SRU is an easy to set up endpoint for Clarin FCS 2.0 (Federated Content Search) based on the Manatee-open corpus search engine and developed and maintained by the Institute of the Czech National Corpus.
- Full support for the FCQ-QL query language
- definable mapping between FCQ-QL layers and Manatee-open positional attributes
- Level 1 support for basic search via CQL (Context Query Language)
- simultaneous search in multiple defined corpora
- (optional) backlinks to respective concordances in KonText
- a working Linux server with installed Manatee-open library
- Redis database
- Go language compiler and tools
- (optional) an HTTP proxy server (Nginx, Apache, ...)
- Install
Golanguage environment, either via a package manager or manually from Go download page- make sure
/usr/local/go/binand~/go/binare in your$PATHso you can run any installed Go tools without specifying a full path
- make sure
- Install Manatee-open from the download page. No specific language bindings are required.
configure --with-pcre --disable-python && make && sudo make install && sudo ldconfig
- Get MQuery-SRU sources (
git clone --depth 1 https://github.com/czcorpus/mquery-sru.git) - Run
./configure - Run
make - Run
make install- the application will be installed in
/opt/mquery-sru - for data and registry,
/var/opt/corpora/dataand/var/opt/corpora/registrydirectories will be created - systemd services
mquery-sru-server.serviceandmquery-sru-worker-all.targetwill be created
- the application will be installed in
- Copy at least one corpus and its configuration (registry) into respective directories (
/var/opt/corpora/data,/var/opt/corpora/registry) - Update corpora entries in
/opt/mquery-sru/conf.jsonfile to match your installed corpora - start the service:
systemctl start mquery-sru-serversystemctl start mquery-sru-worker-all.target
In most cases, it is not recommended to expose the server directly to the Internet. It is therefore advisable to put the service behind an HTTP proxy. E.g. in Nginx, the configuration may look like this:
location /mquery-fcs/ {
proxy_pass http://127.0.0.1:8080/;
proxy_set_header Host $http_host;
proxy_redirect off;
proxy_read_timeout 30;
proxy_set_header X-Forwarded-For $remote_addr;
proxy_set_header X-Forwarded-Proto $scheme;
}
It's important to understand that endpoints experiencing low traffic can still benefit from having multiple workers. Specifically, if an endpoint is configured to search across multiple corpora, MQuery-SRU can leverage these workers to execute searches in parallel. This approach can significantly reduce the response time by querying all configured corpora simultaneously, thereby improving efficiency even under conditions of minimal load.
To run the endpoint, you need at least
- to configure listening address and port
- defined path to your Manatee corpora registry (= configuration) files
- defined corpora along with:
- positional attributes to be exposed and also layer names they belong to
- mapping of FCS-QL's
withinstructures (s,sentence,petc.) to your specific corpora structures
- address of your Redis service plus a number of database to be used for passing queries and results around
See configuration reference and/or conf.sample.json for detailed info.
This applies in case make install is not used.
(Here we assume the service will run with user www-data)
Create a directory for logging (e.g. /var/log/mquery-sru) and set proper permissions for www-data to be able to write there.
You can use predefined systemd files from /scripts/systemd/*. Copy (or link) them to /etc/systemd/system and then run:
systemctl enable mquery-sru-server.service
systemctl enable mquery-sru-worker-all.target
Now you can try to run the service:
systemctl start mquery-sru-server
systemctl start mquery-sru-worker-all.target
A CNC instance of MQuery-SRU is running as one of the endpoints for Clarin Content Search page.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent