= Whistlepig
Whistlepig is a minimalist realtime full-text search index. Its goal is
to be as small and maintainable as possible, while still remaining
useful, performant and scalable to large corpora. If you want realtime
full-text search without the frills, Whistlepig may be for you.
Whistlepig is written in ANSI C99. It currently provides a C API and Ruby
bindings.
Latest version: 0.12, released 2012-06-09.
Status: beta
News: http://all-thing.net/label/whistlepig/
Homepage: http://masanjin.net/whistlepig/
Bug reports: http://github.com/wmorgan/whistlepig/issues
= Getting it
Tarball: http://masanjin.net/whistlepig/whistlepig-0.12.tar.gz
Rubygem: gem install whistlepig
Git: git clone git://github.com/wmorgan/whistlepig.git
= Realtime search
Roughly speaking, realtime search means:
- documents are available to to queries immediately after indexing, without any
reindexing or index merging;
- later documents are more important than earlier documents.
Whistlepig takes these principles at face value.
- It only returns documents in the reverse (LIFO) order to which they were
added, and performs no ranking, reordering, or scoring.
- It only supports incremental indexing. There is no notion of batch indexing
or index merging.
- It does not support document deletion or modification (except in the
special case of labels; see below).
- It only supports in-memory indexes.
Features that Whistlepig does provide:
- Incremental indexing. Updates to the index are immediately available to
readers.
- Fielded terms with arbitrary fields.
- A full query language and parser with conjunctions, disjunctions, phrases,
negations, grouping, and nesting.
- Labels: arbitrary tokens which can be added to and removed from documents
at any point, and incorporated into search queries.
- Early query termination and resumable queries.
- A tiny, < 3 KLOC ANSI C99 implementation.
== Benchmarks
On my not-particularly-new Linux desktop, I can index 8.5 MB/s of text
data per process, including some minor parsing.
Index sizes are roughly 50% of the original corpus size, e.g. the 1.4gb
Enron email corpus (http://cs.cmu.edu/~enron/) is 753mb in the index.
Query performance is entirely dependent on the queries and the index
size. Run the benchmark-queries to see some examples.
== Synopsis (using Ruby bindings)
require 'rubygems'
require 'whistlepig'
include Whistlepig
index = Index.new "index"
entry1 = Entry.new
entry1.add_string "body", "hello there bob"
docid1 = index.add_entry entry1 # => 1
entry2 = Entry.new
entry2.add_string "body", "goodbye bob"
docid2 = index.add_entry entry2 # => 2
q1 = Query.new "body", "bob"
results1 = index.search q1 # => [2, 1]
q2 = q1.and Query.new("body", "hello")
results2 = index.search q2 # => [1]
index.add_label docid2, "funny"
q3 = Query.new "body", "bob ~funny"
results3 = index.search q3 # => [2]
entry3 = Entry.new
entry3.add_string "body", "hello joe"
entry3.add_string "subject", "what do you know?"
docid3 = index.add_entry entry3 # => 3
q4 = Query.new "body", "subject:know hello"
results4 = index.search q4 # => [3]
== Concurrency
Whistlepig supports multi-process concurrency. Multiple reader and
writer processes can access the same index without mangling data.
Internally, Whistlepig uses pthread read-write locks to synchronize
readers and writers. This allows multiple concurrent readers but only a
single writer.
While this locking approach guarantees index correctness, it decreases
read and write performance when one or more writers exist. Systems with
high write loads may benefit from sharding documents across independent
indexes rather than sending everything to the same index.
== Design tradeoffs
I have generally erred on the side of maintainable code at the expense
of speed. Simpler implementations have been preferred over more complex,
faster versions. If you ever have to modify Whistlepig to suit your
needs, you will appreciate this.
== Bug reports
Please file bugs here: https://github.com/wmorgan/whistlepig/issues
Please send comments to: [email protected].
