I recently looked at my Gmail inbox and noticed that I have well over 50k emails, taking up about 12GB of space but there is no good way to tell what emails take up space, who sent them to, who emails me, etc
Goal of this tutorial is to load an entire Gmail inbox into Elasticsearch using bulk indexing and then start querying the cluster to get a better picture of what's going on.
Related tutorial: Index and Search Hacker News using Elasticsearch and the HN API
Set up Elasticsearch and make sure it's running at http://localhost:9200
I use Python and Tornado for the scripts to import and query the data. Run pip install tornado to install Tornado.
First, go here and download your Gmail mailbox, depending on the amount of emails you have accumulated this might take a while.
The downloaded archive is in the mbox format and Python provides libraries to work with the mbox format so that's easy.
The overall program will look something like this:
mbox = mailbox.UnixMailbox(open('emails.mbox', 'rb'), email.message_from_file)
for msg in mbox:
item = convert_msg_to_json(msg)
upload_item_to_es(item)
print "Done!"The full Python code here: src/update.py
First, we got to turn the mbox format messages into JSON so we can insert it into Elasticsearch. Here is some sample code that was very useful when it came to normalizing and cleaning up the data.
A good first step:
def convert_msg_to_json(msg):
result = {'parts': []}
for (k, v) in msg.items():
result[k.lower()] = v.decode('utf-8', 'ignore')Additionally, you also want to parse and normalize the From and To email addresses:
for k in ['to', 'cc', 'bcc']:
if not result.get(k):
continue
emails_split = result[k].replace('\n', '').replace('\t', '').replace('\r', '').replace(' ', '').encode('utf8').decode('utf-8', 'ignore').split(',')
result[k] = [ normalize_email(e) for e in emails_split]
if "from" in result:
result['from'] = normalize_email(result['from'])Elasticsearch expects timestamps to be in microseconds so let's convert the date accordingly
if "date" in result:
tt = email.utils.parsedate_tz(result['date'])
result['date_ts'] = int(calendar.timegm(tt) - tt[9]) * 1000We also need to split up and normalize the labels
labels = []
if "x-gmail-labels" in result:
labels = [l.strip().lower() for l in result["x-gmail-labels"].split(',')]
del result["x-gmail-labels"]
result['labels'] = labelsEmail size is also interesting so let's break that out
parts = json_msg.get("parts", [])
json_msg['content_size_total'] = 0
for part in parts:
json_msg['content_size_total'] += len(part.get('content', ""))The most simple aproach is a PUT request per item:
def upload_item_to_es(item):
es_url = "http://localhost:9200/gmail/email/%s" % (item['message-id'])
request = HTTPRequest(es_url, method="PUT", body=json.dumps(item), request_timeout=10)
response = yield http_client.fetch(request)
if not response.code in [200, 201]:
print "\nfailed to add item %s" % item['message-id']
However, Elasticsearch provides a better method for importing large chunks of data: bulk indexing
Instead of making a HTTP request per document and indexing individually, we batch them in chunks of eg. 1000 documents and then index them.
Bulk messages are of the format:
cmd\n
doc\n
cmd\n
doc\n
...
where cmd is the control message for each doc we want to index.
For our example, cmd would look like this:
cmd = {'index': {'_index': 'gmail', '_type': 'email', '_id': item['message-id']}}`
The final code looks something like this:
upload_data = list()
for msg in mbox:
item = convert_msg_to_json(msg)
upload_data.append(item)
if len(upload_data) == 100:
upload_batch(upload_data)
upload_data = list()
if upload_data:
upload_batch(upload_data)and
def upload_batch(upload_data):
upload_data_txt = ""
for item in upload_data:
cmd = {'index': {'_index': 'gmail', '_type': 'email', '_id': item['message-id']}}
upload_data_txt += json.dumps(cmd) + "\n"
upload_data_txt += json.dumps(item) + "\n"
request = HTTPRequest("http://localhost:9200/_bulk", method="POST", body=upload_data_txt, request_timeout=240)
response = http_client.fetch(request)
result = json.loads(response.body)
if 'errors' in result:
print result['errors']After indexing all your emails, we can start running queries.
If you want to search for emails from the last 6 months, you can use the range filter and search for gte the current time (now) minus 6 month:
curl -XGET 'http://localhost:9200/gmail/email/_search?pretty' -d '{
"filter": { "range" : { "date_ts" : { "gte": "now-6M" } } } }
'
or you can filter for all emails from 2014 by using gte and lt
curl -XGET 'http://localhost:9200/gmail/email/_search?pretty' -d '{
"filter": { "range" : { "date_ts" : { "gte": "2013-01-01T00:00:00.000Z", "lt": "2014-01-01T00:00:00.000Z" } } } }
'
You can also quickly query for certain fields via the q parameter. This example shows you all your Amazon shipping info emails:
curl "localhost:9200/gmail/email/_search?pretty&q=from:[email protected]"
Aggregation queries let us bucket data by a given key and count the number of messages per bucket. For example, number of messages grouped by recipient:
curl -XGET 'http://localhost:9200/gmail/email/_search?pretty&search_type=count' -d '{
"aggs": { "emails": { "terms" : { "field" : "to", "size": 10 }
} } }
'
Result:
"aggregations" : {
"emails" : {
"buckets" : [ {
"key" : "[email protected]",
"doc_count" : 1920
}, { "key" : "[email protected]",
"doc_count" : 1326
}, { "key" : "[email protected]",
"doc_count" : 263
}, { "key" : "[email protected]",
"doc_count" : 232
}
...
]
}
This one gives us the number of emails per label:
curl -XGET 'http://localhost:9200/gmail/email/_search?pretty&search_type=count' -d '{
"aggs": { "labels": { "terms" : { "field" : "labels", "size": 10 }
} } }
'
Result:
"hits" : {
"total" : 51794,
},
"aggregations" : {
"labels" : {
"buckets" : [ {
"key" : "important",
"doc_count" : 15430
}, { "key" : "github",
"doc_count" : 4928
}, { "key" : "sent",
"doc_count" : 4285
}, { "key" : "unread",
"doc_count" : 510
},
...
]
}
Use a date histogram you can also count how many emails you sent and received per year:
curl -s "localhost:9200/gmail/email/_search?pretty&search_type=count" -d '
{ "aggs": {
"years": {
"date_histogram": {
"field": "date_ts", "interval": "year"
}}}}
'
Result:
"aggregations" : {
"years" : {
"buckets" : [ {
"key_as_string" : "2004-01-01T00:00:00.000Z",
"key" : 1072915200000,
"doc_count" : 585
}, {
...
}, {
"key_as_string" : "2013-01-01T00:00:00.000Z",
"key" : 1356998400000,
"doc_count" : 12832
}, {
"key_as_string" : "2014-01-01T00:00:00.000Z",
"key" : 1388534400000,
"doc_count" : 7283
} ]
}
- more interesting queries
- schema tweaks
- multi-part message parsing
- blurb about performance
- ...
Open pull requests, issues or email me at [email protected]