line-parser

<dependency>
    <groupId>com.github.marschall</groupId>
    <artifactId>line-parser</artifactId>
    <version>0.5.0</version>
</dependency>

An mmap() based line parser for cases when:

• the start byte position of a line in the file is required
• the length in bytes of a line is required
• only a few character of every line are required

In these cases this library can theoretically be more efficient than BufferedReader because:

• the copy operations of buffered IO are avoided
• the allocation and resizing of an intermediate StringBuffer is avoided
• the allocation of the final String is avoided, only the required substrings are allocated

The performance may still be slower than a than BufferedReader based approach but it should consume much less memory bandwidth and produce only a fraction of the garbage.

As this project gives you a CharSequence instead of a String you may want to have a look at the charsequences which gives you some the String convenience methods while avoiding allocation.

Misc

• the main parsing loop is likely to benefit from on-stack replacement (OSR)
• if you're using UTF-8 with a BOM then the BOM is returned as well
• if you're using UTF-16 with a BOM then the BOM is returned as well
• the library runs on Java 8 but is also a Java 9 module that only requires the jdk.unsupported module besides the java.base module

Usage

LineParser parser = new LineParser();
parser.forEach(path, cs, (line) -> {
  System.out.printf("[%d,%d]%s%n", line.getOffset(), line.getLength(), line.getContent());
});