The data that comes back from GraphQL is not ideal for input to Elastic Search which is designed for full-text searching, which means it loses context when searching arrays. Consider this response from some fictional GraphQL API.
[{
"franchise": {
"name": "rambo",
"episodes": [{
"name": "first blood",
"year": 1982,
"actors": [{
"dob": "1946-07-06",
"name": "Sylvester Stallone",
"nicknames": ["sly"]
}, {
"name": "Jack Starrett"
}, {
"name": "Richard Crenna",
"nationality": "American"
}]
}, {
"name": "first blood part 2",
"year": 1985,
"actors": [{
"dob": "1946-07-06",
"name": "Sylvester Stallone",
"nicknames": ["sly"]
}, {
"name": "Richard Crenna",
"nationality": "American"
}]
}]
}
}, {
"franchise": {
"name": "terminator",
"episodes": [{
"name": "the terminator",
"year": 1984,
"actors": [{
"dob": "1947-07-30",
"name": "Arnold Schwarzenegger",
"nicknames": ["arnie"]
}, {
"name": "Linda Hamilton"
}]
}, {
"name": "terminator 2: judgement day",
"year": 1985,
"actors": [{
"dob": "1947-07-30",
"name": "Arnold Schwarzenegger",
"nicknames": ["arnie", "schwarzie"]
}, {
"name": "Michael Biehn"
}]
}]
}
}];
To make a success of this in Elastic search we need to turn it back into a tabular format, a little like you might get back from a complex SQL join. Something like this.
That’s a little more complex than simple object flattening, as we need to create a duplicate row each time there is an array element, and of course, these arrays might occur at any level.
The code
const blowup = ({ ob, sep = '_', cloner = item => JSON.parse(JSON.stringify(item)) }) => {
// initial value must be an array of objects
//if(!Array.isArray(ob)) ob = [ob];
// recursive piece
const makeRows = (sob, rows = [], currentKey = '', cob = {}) => {
// ignore undefined or null items
if (typeof sob === typeof undefined || sob === null) {
return rows;
} else if (Array.isArray(sob)) {
// going to work through an array creating 1 row for each element
// but without adding to the current key
// make deep clone of current object
sob.forEach((f, i) => {
// make clone of what we have so far to replicate across
const clob = cloner(cob);
// the first element updates an existing row
// subsequent elements add to the number of rows
if (i) {
rows.push(clob);
} else {
rows[rows.length ? rows.length - 1 : 0] = clob;
}
// recurse for each element
makeRows(f, rows, currentKey, clob);
});
} else if (typeof (sob) === 'object' && !(sob instanceof Date)) {
Object.keys(sob).forEach((k, i) => {
// add to the key, but nothing to the accumulating object
makeRows(sob[k], rows, currentKey ? currentKey + sep + k : k, cob);
});
} else {
// its a natural value
if (cob.hasOwnProperty('currentKey')) {
// something has gone wrong here - show should probably be a throw
console.log('attempt to to overwrite property', cob, currentKey, 'row', rows.length);
} else {
cob[currentKey] = sob;
}
}
return rows;
};
// do the work - the input data should be a single item in an array of objects
return Array.prototype.concat.apply([], ob.map(f => makeRows(f)));
};
That will give this result
[{
"franchise_name": "rambo",
"franchise_episodes_name": "first blood",
"franchise_episodes_year": 1982,
"franchise_episodes_actors_dob": "1946-07-06",
"franchise_episodes_actors_name": "Sylvester Stallone",
"franchise_episodes_actors_nicknames": "sly"
}, {
"franchise_name": "rambo",
"franchise_episodes_name": "first blood",
"franchise_episodes_year": 1982,
"franchise_episodes_actors_name": "Jack Starrett"
}, {
"franchise_name": "rambo",
"franchise_episodes_name": "first blood",
"franchise_episodes_year": 1982,
"franchise_episodes_actors_name": "Richard Crenna",
"franchise_episodes_actors_nationality": "American"
}, {
"franchise_name": "rambo",
"franchise_episodes_name": "first blood part 2",
"franchise_episodes_year": 1985,
"franchise_episodes_actors_dob": "1946-07-06",
"franchise_episodes_actors_name": "Sylvester Stallone",
"franchise_episodes_actors_nicknames": "sly"
}, {
"franchise_name": "rambo",
"franchise_episodes_name": "first blood part 2",
"franchise_episodes_year": 1985,
"franchise_episodes_actors_name": "Richard Crenna",
"franchise_episodes_actors_nationality": "American"
}, {
"franchise_name": "terminator",
"franchise_episodes_name": "the terminator",
"franchise_episodes_year": 1984,
"franchise_episodes_actors_dob": "1947-07-30",
"franchise_episodes_actors_name": "Arnold Schwarzenegger",
"franchise_episodes_actors_nicknames": "arnie"
}, {
"franchise_name": "terminator",
"franchise_episodes_name": "the terminator",
"franchise_episodes_year": 1984,
"franchise_episodes_actors_name": "Linda Hamilton"
}, {
"franchise_name": "terminator",
"franchise_episodes_name": "terminator 2: judgement day",
"franchise_episodes_year": 1985,
"franchise_episodes_actors_dob": "1947-07-30",
"franchise_episodes_actors_name": "Arnold Schwarzenegger",
"franchise_episodes_actors_nicknames": "arnie"
}, {
"franchise_name": "terminator",
"franchise_episodes_name": "terminator 2: judgement day",
"franchise_episodes_year": 1985,
"franchise_episodes_actors_dob": "1947-07-30",
"franchise_episodes_actors_name": "Arnold Schwarzenegger",
"franchise_episodes_actors_nicknames": "schwarzie"
}, {
"franchise_name": "terminator",
"franchise_episodes_name": "terminator 2: judgement day",
"franchise_episodes_year": 1985,
"franchise_episodes_actors_name": "Michael Biehn"
}]
We can take that one stage further and convert the result into a tabular format (ideal for writing to a sheet)
const blownupToTable = ({ blownup, sorter = mentions => Object.keys(mentions).sort((a, b) => a - b) }) => {
// collect all the property names
const mentions = blownup.reduce((p, c) => {
Object.keys(c).forEach((k, i) => {
p[k] = i;
});
return p;
}, {});
// make that into a header row
const headerRow = sorter(mentions);
// now add the rows after the header
// & we dont really like undefined in sheets, so replace with null.
return [headerRow]
.concat(blownup.map(row => headerRow.map(h => typeof row[h] === typeof undefined ? null : row[h])));
}
Which gives this result, with the header row in the first element
[ ["franchise_name", "franchise_episodes_name", "franchise_episodes_year", "franchise_episodes_actors_dob", "franchise_episodes_actors_name", "franchise_episodes_actors_nicknames", "franchise_episodes_actors_nationality"], ["rambo", "first blood", 1982, "1946-07-06", "Sylvester Stallone", "sly", null], ["rambo", "first blood", 1982, null, "Jack Starrett", null, null], ["rambo", "first blood", 1982, null, "Richard Crenna", null, "American"], ["rambo", "first blood part 2", 1985, "1946-07-06", "Sylvester Stallone", "sly", null], ["rambo", "first blood part 2", 1985, null, "Richard Crenna", null, "American"], ["terminator", "the terminator", 1984, "1947-07-30", "Arnold Schwarzenegger", "arnie", null], ["terminator", "the terminator", 1984, null, "Linda Hamilton", null, null], ["terminator", "terminator 2: judgement day", 1985, "1947-07-30", "Arnold Schwarzenegger", "arnie", null], ["terminator", "terminator 2: judgement day", 1985, "1947-07-30", "Arnold Schwarzenegger", "schwarzie", null], ["terminator", "terminator 2: judgement day", 1985, null, "Michael Biehn", null, null] ]
Putting it together
Assuming the JSON data earlier in the article is in a variable called films, here’s the whole thing
const objectUnnest = ({ ob, sep, cloner, sorter }) => {
const blownup = blowup({ ob, sep, cloner});
const result = blownupToTable ({ blownup, sorter });
return result;
}
console.log(objectUnnest({ ob: films }));
You can find an Apps Script version of this here.
Since G+ is closed, you can now star and follow post announcements and discussions on Github, here