mwmitchell / flux Goto Github PK

In order to make Flux simpler to get started with, it would be useful to have examples that show how to query Solr using Flux. Some of the Solr components that would be nice to see could be some of these:

• https://docs.lucidworks.com/display/solr/The+Extended+DisMax+Query+Parser
• https://docs.lucidworks.com/display/solr/Faceting
• https://docs.lucidworks.com/display/solr/The+Stats+Component

These examples could be inlined with the main README file or as a separate document.

Hi,

I've written a lot of integration tests for flux, as well as some comments on what people should expect when querying. What's the contribution procedure? I normally use git-flow, so I'm building them on a development branch on my fork, but flux seems to keep only master (and dev is out of date).

It took a little while for me to realise that the flux.cloud namespace was added after the current release. Is there any chance of that release being made? I'm keen to use Flux to connect to a SolrCloud cluster!

Currently, Flux lacks documentation on its code/functions. Because of this, there's no way to build docs dynamically. Any public function/namespace within Flux should be documented with a doc string. Something like marginalia [1] could then be used for publishing docs.

1. https://github.com/gdeer81/marginalia

Just got to the part I try to connect to a real solr and it seems there are few problems the midje test not digged into. I'm on it and will let you know when everything is fixed and tested against solr.

ToDo:

• Write tests for http connection
• Fix all errors in http connection
• Write tests for embedded connection
• Fix all errors in embedded connection
• Write tests for cloud connection
• Fix all errors in cloud connection
• Update readme

I'm currently not sure if the embedded version is of use? https://cwiki.apache.org/confluence/display/solr/EmbeddedSolr state:

The concept of "Embedded Solr" came about in 2007 when some Java developers wanted to get the goodies Solr added on top of the Lucene Java API, but without running a separate Java process.

This wiki page originally served as an example of how, by directly compiling against the classes provided in the Solr WAR, you could add code directly to an existing Java application to do the necessary setup to initialize a SolrCore, add documents to the index, and execute queries.

Those examples are still available in the historic versions of this page, but they are no longer recommended (not that they ever really were).

The simplest, safest, way to use Solr is via Solr's standard HTTP interfaces. Embedding Solr is less flexible, harder to support, not as well tested, and should be reserved for special circumstances.

SolrJ offers more - leverage more of the methods available - https://cwiki.apache.org/confluence/display/solr/Using+SolrJ

As you know I am playing with flux currently without much knowledge of solr. Inspired by some sql libraries what do you think about a more clojure way building queries?

What I've done so far:

(q (or {:en 1} 
       (and {"de" 2}
            (+ :foo)
            (- "test")) 
       {:fr (not 4 "bar")}))
;; => "q=(en:1 || (de:2 &&  +foo &&  -test) || fr ! 4 && fr ! bar)"

;; Do you prefer (implemented both, nothing more to decide here)
(q {"foo" [:* 4]})
;; or
(q {"foo" (range :* 4)})
;; => "q=foo:[* TO 4]"

I'm not sure if those prohibits and requires are correct this way.

As said I start using solr now and not build many queries before. It would help me to add cases to this query builder when you show me some sample query strings.

Not sure if it's all if I go through examples of solr as alternative

It would be great to see a SolrServer type hint on this call to .query. Profiling shows that a significant amount of CPU is being spent on reflection against the solr-server object. Obviously, type hints against other method calls would also be of benefit, but that's the big one we're seeing.

Would a pull request be helpful?

Commit 62ae93e introduced a regression when converting NamedList objects. In previous version, covert-named-list returned a map with string keys converted to keywords. The new ->clojure multi method will return a vector of 2-element vectors when called on a NamedList (string keys not converted). Fix available in pull request [#4].