I am pretty new to CouchDB, Map/Reduce and NoSQL in general.
I was hoping if you guys could guide me on how I would implement very basic search on my Node.js/ CouchDB application.
I am looking at searching for certain text elements within the CouchDB documents.
Most of my documents within couch are of a similar format mentioned below :
{
"_id": "b001729a2c5cf4100ea50a71ec04e20b",
"_rev": "1-8928ea122d80fd5c1da1c22dfc6ce46e",
"Approved To Publish": "false",
"Article Body": "has retained the Carbon Trust Standard for energy and carbon management.Following a demanding audit process that took place in December 2012, was awarded recertification for the Carbon Trust Standard.This certificate is a mark of excellence in recognition of how well we measure, manage and reduce our carbon emissions.",
"Headline": "Delight retains Carbon Trust Standard"
}
My search keys would be for eg 'Carbon trust', 'emissions', 'excellence recognition' etc.
What I have so for is a temporary map function which I use in the request body of my Node.js application in a POST request, but I am sure its not the right approach, and I would expect it to be a stored view in CouchDB.
My Map function:
function (doc) {
if ('Headline' in doc) {
if (doc['Article Body'].search(/" + req.body.jsonValue + "/i) !== -1
|| doc.Headline.search(/" + req.body.jsonValue + "/i) !== -1) {
var key = doc.Headline,
value = doc;
emit(key, value);
}
}
}
Please let me know on what I need to do to improve my approach or let me know if things are unclear.
Regards
A list function has access to querystring values, so you can just add one that you use in conjunction with your view.
function (doc) {
if ("Headline" in doc) {
emit(doc.Headline, doc);
}
}
function (head, req) {
var rows = [],
regex = new RegExp(req.query.search, "i"), // use the querystring param to create the RegExp object
row;
while (row = getRow()) {
if (row.value.Headline.search(regex) > -1 || row.value["Article Body"].search(regex)) {
rows.push(row);
}
}
// I'm just mocking up what a view's output would look like (not required)
send(JSON.stringify({
total_rows: rows.length,
offset: 0,
rows: rows
}));
}
Of course you could modify the list function to emit data in chunks, rather than all at once, but this should give you an idea of what a solution like this looks like.
You could instead build a view where keys are keywords (or alternatively any words) and values are _ids.
The drawback is that this view could potentially become very large. CouchDB experts may have a better solution to what is I suppose a typical problem.
Naive example1:
function(doc) {
if ('Headline' in doc) {
for (key in doc.Headline.split(' ')) {
emit(key, doc._id)
}
}
if ('Article Body' in doc) {
for (key in doc['Article Body'].split(' ')) {
emit(key, doc._id)
}
}
}
Then you would query it with /app/_design/app/_view/search?key="keyword" for instance.
1: you would actually need to normalize case, remove punctuation, common words like a, the, of and more…