Sequences In-Depth

Learning Objective
You will learn in detail how to optimize the usage of sequences dependent on your needs.
Difficulty
Advanced
Duration
20 min
Prerequisites
Sequences

Sequences, particularly Strings, are fundamental in SeqAn. You learned already how to use the default implementation of strings and how to easily work with them. In the most cases the default string specialization is well suited as well as the default behavior for capicity changes. Nevertheless, sometimes you might want to change the default behavior for efficiency reasons and adjust it to your specific needs.

String Specializations

In this section you will learn about the different string specializations and when to use them.

The user can specify the kind of string that should be used in an optional second template argument of String.

String<Dna>           dnaSeq1; // The default string implementation: Alloc
String<Dna, Alloc<> > dnaSeq2; // The same as above

In most cases, the implementation Alloc String (the default when using a String<T>) is the best choice. Exceptions are when you want to process extremely large strings that are a bit larger than the available memory (consider Alloc String) or much larger so most of them are stored on the hard disk and only parts of them are loaded in main memory (consider External String).

The following list describes in detail the different specializations:

Specialization Alloc String
  • Description Expandable string that is stored on the heap.
  • Applications The default string implementation that can be used for general purposes.
  • Limitations Changing the capacity can be very costly since all values must be copied.
Specialization Array String
  • Description Fast but non-expandable string.Fast storing of fixed-size sequences.
  • Limitations Capacity must already be known at compile time. Not suitable for storing large sequences.
Specialization Block String
  • Description String that stores its sequence characters in blocks.
  • Applications The capacity of the string can quickly be increased. Good choice for growing strings or stacks.
  • Limitations Iteration and random access to values is slightly slower than for Alloc String.
Specialization Packed String
  • Description A string that stores as many values in one machine word as possible.
  • Applications Suitable for storing large strings in memory.
  • Limitations Slower than other in-memory strings.
Specialization External String
  • Description String that is stored in secondary memory.
  • Applications Suitable for storing very large strings (>2GB). Parts of the string are automatically loaded from secondary memory on demand.
  • LimitationsApplications Slower than other string classes.
Specialization CStyle String
  • Description Allows adaption of strings to C-style strings.
  • Applications Used for transforming other String classes into C-style strings (i.e. null terminated char arrays). Useful for calling functions of C-libraries.
  • Limitations Only sensible if value type is char or wchar_t.
// String with maximum length 100.
String<char, Array<100> > myArrayString;
// String that takes only 2 bits per nucleotide.
String<Dna, Packed<> > myPackedString;

Overflow Strategies

The following section will describe how you can improve capacity changes for your sequences.

Each sequence object has a capacity, i.e. the reserved space for this object. The capacity can be set explicitly by functions such as reserve or resize. It can also bet set implicitly by functions like append, assign, insert or replace, if the operation’s result exceeds the length of the target sequence.

If the current capacity of a sequence is exceeded by chaning the length, we say that the sequence overflows. There are several overflow strategies that determine what actually happens when a string should be expanded beyond its capacity. The user can specify this for a function call by additionally handing over a tag. If no overflow strategy is specified, a default overflow strategy is selected depending on the type of the sequence.

The following overflow strategies exist:

Exact
Expand the sequence exactly as far as needed. The capacity is only changed if the current capacity is not large enough.
Generous
Whenever the capacity is exceeded, the new capacity is chosen somewhat larger than currently needed. This way, the number of capacity changes islimited in a way that resizing the sequence only takes amortized constant time.
Limit
Instead of changing the capacity, the contents are limited to current capacity. All values that exceed the capacity are lost.
Insist
No capacity check is performed, so the user has to ensure that the container’s capacity is large enough.

The next example illustrates how the different strategies could be used:

String<Dna> dnaSeq;
// Sets the capacity of dnaSeq to 5.
resize(dnaSeq, 4, Exact());
// Only "TATA" is assigned to dnaSeq, since dnaSeq is limited to 4.
assign(str, "TATAGGGG", Limit());
std::cout << dnaSeq << std::endl;
// Use the default expansion strategy.
append(dnaSeq, "GCGCGC");
std::cout << dnaSeq << std::endl;
TATA
TATAGCGCGC

Workshop Assignment 1

Type
Review
Objective
Build a string of Dna (default specialization) and use the function appendValue to append a million times the nucleotide ‘A’. Do it both using the overflow strategy Exact and Generous. Measure the time for the two different strategies.
Solution

Click more... to see the solution.

#include <iostream>
#include <seqan/sequence.h>
#include <seqan/file.h>

#include <time.h>  

using namespace seqan;

int main()
{
    unsigned num = 1000000;
    time_t start;

    String<Dna> str;
    clear(str);
    start = time (NULL); 
    for (unsigned i = 0; i < num; ++i){

        appendValue(str, 'A', Exact());
    }
    std::cout << "Strategy Exact() took: " << time(NULL) - start << " s\n\n";

    clear(str);
    start = time(NULL);
    for (unsigned i = 0; i < num; ++i){

        appendValue(str, 'A', Generous());
    }
    std::cout << "Strategy Generous() took: " << time(NULL) - start << " s\n\n";

    return 0;
}
comments powered by Disqus