In a Redis database I have a number of hashes corresponding to "story" objects.
I have an ordered set stories containing all keys of the above (the stories) enabling convenient retrieval of stories.
I now want to store arbitrary emoticons (ie. the Unicode characters corresponding to "smiley face" etc) with stories as "user emotions" corresponding to the emotion the story made the user feel.
I am thinking of:
creating new hashes called emotions containing single emoticons (one per emotion expressed)
creating a hash called story-emotions that enables efficient retrieval of and counting of all the emotions associated with a story
creating another new hash called user-story-emotions mapping user IDs to items in the story-emotion hash.
Typical queries will be:
retrieve all the emotions for a story for the current user
retrieve the count of each kind of emotion for the 50 latest stories
Does this sound like a sensible approach?
Very sensible, but I think I can help make it even more so.
To store the emoticons dictionary, use two Hashes. The first, lets call it emoticon-id should have a field for each emoticon expressed. The field name is the actual Unicode sequence and the value is a unique integer value starting from 0, and increasing for each new emoticon added.
Another Hash, id-emoticon, should be put in place to do the reverse mapping, i.e. from field names that are ids to actual Unicode values.
This gives you O(1) lookups for emoticons, and you should also consider caching this in your app.
To store the user-story-emotions data, look into Redis' Bitmaps. Tersely, use the emoticon id as index to toggle the presence/lack of it by that user towards that story.
Note that in order to keep things compact, you'll want popular emotions to have low ids so your bitmaps remain a small as possible.
To store the aggregative story-emotions, the Sorted Set would be a better option. Elements can be either id or actual unicode, and the score should be the current count. This will allow you to fetch the top emoticons (ZREVRANGEBYSCORE) and/or page similarly to how you're doing with the recent 50 stories (I assume you're using the stories Sorted Set for that).
Lastly, when serving the second query, use pipelining or Lua scripting when fetching the bulk of 50 story-emotion counter values in order get more throughput and better concurrency.
Related
I have a web / mobile application that should display an infinite scroll view (the continuation of the list of items is loaded periodically in a dynamic way) with items where each of the items have a weight, the bigger is the weight in comparison to the weights of other items the higher should be the chances/probability to load the item and display it in the list for the users, the items should be loaded randomly, just the chances for the items to be in the list should be different.
I am searching for an efficient algorithm / solution or at least hints that would help me achieve that.
Some points worth to mention:
the weight has those boundaries: 0 <= w < infinite.
the weight is not a static value, it can change over time based on some item properties.
every item with a weight higher than 0 should have a chance to be displayed to the user even if the weight is significantly lower than the weight of other items.
when the users scrolls and performs multiple requests to API, he/she should not see duplicate items or at least the chance should be low.
I use a SQL Database (PostgreSQL) for storing items so the solution should be efficient for this type of database. (It shouldn't be a purely SQL solution)
Hope I didn't miss anything important. Let me know if I did.
The following are some ideas to implement the solution:
The database table should have a column where each entry is a number generated as follows:
log(R) / W,
where—
W is the record's weight greater than 0 (itself its own column), and
R is a per-record uniform random number in (0, 1)
(see also Arratia, R., "On the amount of dependence in the prime factorization of a uniform random integer", 2002). Then take the records with the highest values of that column as the need arises.
However, note that SQL has no standard way to generate random numbers; DBMSs that implement SQL have their own ways to do so (such as RANDOM() for PostgreSQL), but how they work depends on the DBMS (for example, compare MySQL's RAND() with T-SQL's NEWID()).
Peter O had a good idea, but had some issues. I would expand it a bit in favor of being able to shuffle a little better as far as being user-specific, at a higher database space cost:
Use a single column, but store in multiple fields. Recommend you use the Postgres JSONB type (which stores it as json which can be indexed and queried). Use several fields where the log(R) / W. I would say roughly log(U) + log(P) where U is the number of users and P is the number of items with a minimum of probably 5 columns. Add an index over all the fields within the JSONB. Add more fields as the number of users/items get's high enough.
Have a background process that is regularly rotating the numbers in #1. This can cause duplication, but if you are only rotating a small subset of the items at a time (such as O(sqrt(P)) of them), the odds of the user noticing are low. Especially if you are actually querying for data backwards and forwards and stitch/dedup the data together before displaying the next row(s). Careful use of manual pagination adjustments helps a lot here if it's an issue.
Before displaying items, randomly pick one of the index fields and sort the data on that. This means you have a 1 in log(P) + log(U) chance of displaying the same data to the user. Ideally the user would pick a random subset of those index fields (to avoid seeing the same order twice) and use that as the order, but can't think of a way to make that work and be practical. Though a random shuffle of the index and sorting by that might be practical if the randomized weights are normalized, such that the sort order matters.
Introduction
My domain has articles, which have a title and text. Each article has revisions (like the SVN concept), so every time it is changed/edited, those changes will be stored as a revision. A revision is composed of changes and the description of those changes
I want to be able to obtain all revisions descriptions at once.
What's the problem?
I'm certain that I would store the revision as a hash in articles:revisions:<id> storing the changes, and the description in it.
What I'm not certain of is how do I get all of the descriptions at once.
I have many options to do this, but none of them convinces me.
Store the revision ids for an article as a set, and use SORT articles:revisions:idSet BY NOSORT GET articles:revisions:*->description. This means that I would store a set for each article. If every article had 50 revisions, and we had 10.000 articles, we would have 500.000 ids stored.
Is this the best way? Isn't this eating up too much RAM?
I have other ideas in mind, but I don't consider them good either.
Iterate from 0 to the last revision's id, doing a HGET for each id using MULTI
Create the idSet for a specific article if it doesn't exist and is request, expire after some time.
Isn't there a way for redis to do a SORT array BY NOSORT GET, with array being an adhoc array in the form of [0, MAX]?
Seems like you have a good solution.
As long as you keep those id numbers less than 10,000 and your sets with less than 512 elements(set-max-intset-entries), your memory consumption will be much lower than you think.
Here's a good explanation of it.
This can be solved in an optimized way using a TRIE or DAWG better than what Redis provides. I don't know your application or other info on your search problem (e.g. construction time, unsuccessful searches, update performance).
If you search much more often than you need to update / insert into your lookup storage, I'd suggest you have a look at DAWGDIC [1] as a library, and construct "search paths" (similar as you already described) using a string format that can be search-completed later:
articleID:revisionID:"changeDescription":"change"
Example (I assume you have one description per revision, and n changes. This isn't clear to me from your question):
1:2:"Some changes":"Added two sentences here, removed one sentence there"
1:2:"Some changes":"Fixed article title"
2:4:"Advertisement changes":"Added this, removed that"
Note: Even though you construct these strings with duplicate prefixes, the DAWG will store them in a very space efficient way (simply put, it will append the right side of the string to the data structure and create a shortcut for the common prefix, see also [2] for a comparison of TRIE data structures).
To list changes of article 1, revision 2, set the common prefix for your lookup:
completer.Start(index, "1:2");
Now you can simple call completer.Next() to lookup a next record that shares the same prefix, and completer.value() to get the record's value. In our example we'll get:
1:2:"Some changes":"Added two sentences here, removed one sentence there"
1:2:"Some changes":"Fixed article title"
Of course you need to parse the strings yourself into your data object.
Maybe that's not what you're looking for and overkill. But it can be a very space and search performance efficient way, if it meets your requirements.
[1] https://code.google.com/p/dawgdic/
[2] http://kmike.ru/python-data-structures/
As an aid to learning objective c/oop, I'm designing an iOS app to store and display periodic bodyweight measurements. I've got a singleton which returns a mutablearray of the shared store of measurement object. Each measurement will have at least a date and a body weight, and I want to be able to add historic measurements.
I'd like to display the measurements in date order. What's the best way to do this? As far as I can see the options are as follows: 1) when adding a measurement - I override addobject to sort the shared store every time after a measurement is added, 2) when retrieving the mutablearray I sort it, or 3) I retrieve the mutablearray in whatever order it happens to be in the shared store, then sort it when displaying the table/chart.
It's likely that the data will be retrieved more frequently than a new datum is added, so option 1 will reduce redundant sorting of the shared store - so this is the best way, yes?
You can use a modified version of (1). Instead of sorting the complete array each time a new object is inserted, you use the method described here: https://stackoverflow.com/a/8180369/1187415 to insert the new object into the array at the correct place.
Then for each insert you have only a binary search to find the correct index for the new object, and the array is always in correct order.
Since you said that the data is more frequently retrieved than new data is added, this seems to be more efficient.
If I forget your special case, this question is not so easy to answer. There are two basic solutions:
Keep array unsorted and when you try to access the element and array is not sorted, then sort it. Let's call it "lazy sorting".
Keep array sorted when inserting elements. Note this is not about appending new element at the end and then sort the whole array. This is about finding where the element should be (binary search) and place it there. Let's call it "sorted insert".
Both techniques are correct and useful and deciding which one is better depends on your use cases.
Example:
You want to insert hundreds of elements into the array, then access the elements, then again insert hundreds of elements, then access. In summary, you will be inserting values in big chunks. In this case, lazy sorting will be better.
You will often insert individual elements and you will access the elements often. Then sorted insert will have better performance.
Something in the middle (between inserting 1 and inserting tens of elements). You probably don't care which one of the methods will be used.
(Note that you can use also specialized structures to keep an array sorted, not based on NSArray, e.g. structures based on a balanced tree, while keeping number of elements in the subtree).
Please comment and critique the approach.
Scenario: I have a large dataset(200 million entries) in a flat file. Data is of the form - a 10 digit phone number followed by 5-6 binary fields.
Every week I will be getting a Delta files which will only contain changes to the data.
Problem : Given a list of items i need to figure out whether each item(which will be the 10 digit number) is present in the dataset.
The approach I have planned :
Will parse the dataset and put it a DB(To be done at the start of the
week) like MySQL or Postgres. The reason i want to have RDBMS in the
first step is I want to have full time series data.
Then generate some kind of Key Value store out of this database with
the latest valid data which supports operation to find out whether
each item is present in the dataset or not(Thinking some kind of a
NOSQL db, like Redis here optimised for search. Should have
persistence and be distributed). This datastructure will be read-only.
Query this key value store to find out whether each item is present
(if possible match a list of values all at once instead of matching
one item at a time). Want this to be blazing fast. Will be using this functionality as the back-end to a REST API
Sidenote: Language of my preference is Python.
A few considerations for the fast lookup:
If you want to check a set of numbers at a time, you could use the Redis SINTER which performs set intersection.
You might benefit from using a grid structure by distributing number ranges over some hash function such as the first digit of the phone number (there are probably better ones, you have to experiment), this would e.g. reduce the size per node, when using an optimal hash, to near 20 million entries when using 10 nodes.
If you expect duplicate requests, which is quite likely, you could cache the last n requested phone numbers in a smaller set and query that one first.
I am trying to implement BM25f scoring system on Lucene. I need to make a few minor changes to the original implementation given here for my needs, I got lost at the part where he gets Average Field Length and document length... Could someone guide me as to how or where I get it from?
You can get field length from TermVector instances associated with documents' fields, but that will increase your index size. This is probably the way to go unless you cannot afford a larger index. Of course you will still need to calculate the average yourself, and store it elsewhere (or perhaps in a special document with a well-known external id that you just update when the statistics change).
If you can store the data outside of the index, one thing you can do is count the tokens when documents are tokenized, and store the counts for averaging. If your document collection is static, just dump the values for each field into a file & process after indexing. If the index needs to get updated with additions only, you can store the number of documents and the average length per field, and recompute the average. If documents are going to be removed, and you need an accurate count, you will need to re-parse the document being removed to know how many terms each field contained, or get the length from the TermVector if you are using that.