Abhijeet Mohapatra

Incremental Maintenance of Aggregate Views

We propose an algorithm called CReaM to incrementally maintain materialized aggregate views with user-defined aggregates in response to changes to the database tables from which the view is derived. CReaM is optimal and guarantees the self-maintainability of aggregate views that are defined over a single database table. For aggregate views that are defined over multiple database tables and do not contain all of the non-aggregated attributes in the database tables, CReaM speeds up the time taken to update a view as compared to prior view maintenance techniques. The speed up in the time taken to update a materialized view with n tuples is either

Rule-Based Exploration of Structured Data in the Browser

tfrac{n}{log n}

Incrementally maintaining run-length encoded attributes in column stores

or logn depending on whether the materialized view is indexed or not. For other types of aggregate views, CReaM updates the view in no more time than that is required by prior view maintenance techniques to update the view.

Optimizing fine grained access control using authorization indexes

We present Dexter, a browser-based, domain-independent structured-data explorer for users. Dexter enables users to explore data from multiple local and Web-accessible heterogeneous data sources such as files, Web pages, APIs and databases in the form of tables. Dexter’s users can also compute tables from existing ones as well as validate the tables (base or computed) through declarative rules. Dexter enables users to perform ad hoc queries over their tables with higher expressivity than that is supported by the underlying data sources. Dexter evaluates a user’s query on the client side while evaluating sub-queries on remote sources whenever possible. Dexter also allows users to visualize and share tables, and export (e.g., in JSON, plain XML, and RuleML) tables along with their computation rules. Dexter has been tested for a variety of data sets from domains such as government and apparel manufacturing. Dexter is available online at http://dexter.stanford.edu.

EFFICIENT DATABASE COMPRESSION

Run-length encoding is a popular compression scheme which is used extensively to compress the attribute values in column stores. Out of order insertion of tuples potentially degrades the compression achieved using run-length encoding and consequently, the performance of reads. The in-place insertions, deletions and updates of tuples into a column store relation with n tuples take O(n) time. The linear cost is typically avoided by amortizing the cost of updates in batches. However, the relation is decompressed and subsequently re-compressed after applying a batch of updates. This leads to added time time complexity. We propose a novel indexing scheme called count indexes that supports O(log n) in-place insertions, deletions, updates and look ups on a run-length encoded sequence with n runs. We also show that count indexes efficiently update a batch of tuples requiring almost a constant time per updated tuple. Additionally, we show that count indexes are optimal. We extend count indexes to support O(log n) updates on bitmapped sequences with n values and adapt them to block-based stores.