Mongoose Middleware Super Query

[that1guy]

Just combing over documentation on your Mongoose Middleware (seriously awesome stuff man!)

Today when searching for items internally in htsApp our Mongo query looks like:

PostModel.find({
                coordinates: {
                    $near: {
                        $geometry: {
                            type: "Point",
                            coordinates: [ userCurrentLongBasedOnIP , userCurrentLatBasedOnIP ]
                        },
                        $maxDistance: maxDistanceAwayFromUser,
                        $minDistance: minDistanceAwayFromUser
                    }
                },
                $or: popularCategories,
                heading: { "$regex": searchTerm, "$options": 'i' }
            }, function (err, hts_results) {
               .
               .
               do something here
               .
               .
            })

Quick breakdown of this query:

1. Simple geospatial Mongo query around the user's latitude, longitude we got from freeGeoIP
2. We define the how far around the user we should look (e.g. If 3Taps had to search 30 miles out to find their first 99 items then that's how far out we search our internal database so that we can merge our internal results with 3Taps response).

2a. The $maxDistance and $minDistance params are increased as the user paginates.

1. $or popularCategories is the result of an algorithm which determined our user's searchTerm, "toolboxes", is mostly found in the STOO and VPAR categories.
2. Today I'm performing a very weak $regex string match (case insensitive) of the user's searchTerm against the heading field. I was forced to do a $regex instead of using Mongo's $text functionality because you cannot combine $text and $near in the same query (see my Stackoverflow: http://stackoverflow.com/questions/25922965/mongodb-text-with-near).

I thought this breakdown might help determine our goal in leveraging your Mongoose Middleware to formulate a geospatial, Elasticsearch-capable, category-intelligent super query! :)

Cheers man!

• Brad

[brozeph]

Working on a geo element to the query string now... it looks like the following:

https://localhost:4043/v1/postings?geo[min]=0&geo[max]=5000&geo[coords]=-122.30257,47.44376

Where:

• geo[min] = meters minimum distance
• geo[max] = meters maximum distance
• geo[coords] = longitude, latitude

In the query results, the geo values appear as follows:

"options": {
    "count": 1000,
    "filters": {},
    "geo": {
      "min": "0",
      "max": "1000",
      "coords": "-122.30257,47.44376"
    },
    "sort": {},
    "start": 0
}

[brozeph]

Implemented a migration to insure index is correct type:

bash init/apply-migrations.sh

[brozeph]

https://localhost:4043/v1/postings?geo[min]=0&geo[max]=5000&geo[coords]=122.3750,37.6189

All postings within 5000 meters of SF

[brozeph]

https://localhost:4043/v1/postings?geo[min]=0&geo[max]=1000&geo[coords]=-122.30257,47.44376&filters[mandatory][contains][heading]=wallet

All postings within 1000 meters of the Seattle airport with a heading that contains the word "wallet"

[brozeph]

Verified - closing

[that1guy]

I'm gonna leave this ticket open as it's giving a pretty large definition of query functionality we're aiming for. Let me know if this ticket too large and I can break into smaller tickets.

• Geo-spacial search - Done
• Field Contains (Text Search) - Done
• Pagination - Done
• Sort by distance from Lat, Long - Done
• Smart Category Targeting (Only search in specified cats) - Done
• Field Fuzzy Text ElasticSearch - In progress

Just an example of what I'm thinking.. totally up to you.

    var params = {
        filters: {
            mandatory: {
                fuzzySearch: {
                    heading: "bike"
                }
            }
        },
        geo: {
           min: {0},
           max: {1000},
           coords: {[
               long,
               lat
           ]}
        },
        sort: {
            distance: {[
               long,
               lat
           ]}
        }
    };

    $http({
        method: 'GET',
        url: ENV.postingAPI + utilsFactory.bracketNotationURL(params)
    }).then(function (response) {
        console.log('here it is', response);
        deferred.resolve(response);
    }, function (err) {
        deferred.reject(err);
    });

[brozeph]

Sort by distance along with distance in meters is complete... when querying by geo, distance appears as an additional field under "geo"

[brozeph]

Example search across multiple categories:

https://localhost:4043/v1/postings?filters[optional][exact][categoryCode]=SAPP,SANT

[that1guy]

31 kinda related. See question.

[that1guy]

This enhancement has been shipped. Built new tickets detailing small bugs within this enhancement. Closing.