JarCastingLog4JCat
Overview
Sometimes you've only got the log files and want or need to trim them, eg. to the last 2 hours programatically. Not a well defined task, but we can have a pretty good guess (unless the log format is plain weird.) As log files can be pretty large, up to several GB in some cases, it would take a long time to scan them line-by-line. However, since all entries within a log file are sorted chronologically from oldest to newest, we can perform a quick binary search.
This is what this does in a nutshell, for LOG4J log files with its log entries formatted according to a certain specified Pattern Layout.
Example:
public class Sandbox {
public static void main(final String[] args) throws IOException {
// Create a new log file trimmer with a specified log entry format.
// Log files with this format look like this
// [DEBUG] 2017-06-22 21:57:53,661 TimeClass - Past
// [INFO ] 2017-06-23 21:57:53,661 TimeClass - Present
// [WARN ] 2017-06-24 21:57:53,661 TimeClass - Future
final Log4JCat cat = Log4J.of("[%-5p] %d %c - %m%n").get();
// Open a stream for the log file.
try (final IRandomAccessInput input = InputFactory.open(new File("~/mylogfile"))) {
// Trim the log file so that it contains only entries more recent than the current date.
// This returns the offset in the file where the entries start.
final long pos = cat.tail(input, Timestamp.now());
// Seek to the starting position.
input.seek(pos);
// And print the all lines from the starting position.
System.out.print(input.readLines());
}
}
}
Installation
Use maven central:
<dependency>
<groupId>com.github.blutorange</groupId>
<artifactId>log4jcat</artifactId>
<version>0.1</version>
</dependency>
Or clone the project and run a maven package.
Performance
I tried it with UTF-8 log files about 1GB in size and it performs pretty much as fast as your hard drive (0.01s-0.1s). Parsing the entire log file entry by entry takes around a minute. I'm pretty sure the log file parser can be optimized, but with a binary search O(log(n)), there's no need: I did not see any significant difference between a 500MB and 1GB file, so I suppose most of the time is taken up by initializing the file read.
Documentation
Trimming
The following methods are available for log file trimming, provided by Log4JCat:
- long find(IRandomAccessInput, long) Takes a log file and a UNIX timestamp. Some log entries lie before the given date, and some lie after the given date. This method finds the first log entry that lies after or on the given date. Call this method twice to perform a head-tail trim.
Appender
There are also some utility methods available in the class Cyperus.
These all take the log file from an Appender or Logger, filter the log entries and write them to the given OutputStream. The Appender must be a FileAppender. RollingFileAppender with multiple older files is supported. When a Logger is passed, it takes all its FileAppenders, filters their log files and writes them to the OutputStream.
- Cyperus.head(Appender, OutputStream, long)
- Cyperus.head(Logger, OutputStream, long)
- Cyperus.trim(Appender, OutputStream, long, long)
- Cyperus.trim(Logger, OutputStream, long, long)
- Cyperus.tail(Appender, OutputStream, long)
- Cyperus.tail(Logger, OutputStream, long)
Timestamp
Timestamp is a small set of utility methods for using other date-time objects. Each method returns a UNIX timestamp.
- Timestamp.now()
- Timestamp.of(java.util.Date)
- Timestamp.of(java.util.Calendar)
- Timestamp.of(java.time.temporal.TemporalAccessor) Such as a ZonedDateTime or Instant.
- Timestamp.of(org.joda.time.ReadableInstant) Such as a org.joda.time.DateTime.
Configuration
At the least, you need to pass a pattern layout to get a builder for configuration.
- Log4J.of(String patternLayout) Returns a new builder.
- Log4J.get() Builds a log file trimmer instance and returns it.
There are more options available you can set via the builder before creating a log file trimmer instance:
- locale(Locale) The locale used when the pattern layout contains abbreviations such as "Jan" or "Monday". Default to Locale.ENGLISH.
- timeZone(TimeZone) The time zone of the dates in the log file. Defaults to UTC.
- threshold(long) In bytes. When the search range has been narrowed down to this threshold, the log file trimmer switches to a linear search algorithm. Defaults to 10000.
Input
To pass the log file, use the InputFactory, which allows you to use both streams, strings, and files as input. They all return an object you can pass to the log file trimming methods provided by Log4JCat.
The following factory methods are available:
- InputFactory.open(File[, Charset|String])
- InputFactory.open(Path[, Charset|String])
- InputFactory.open(RandomAccessFile[, Charset|String])
- InputFactory.open(String) The string represents the log file content.
- InputFactory.open(InputStream[, Charset|String])
- InputFactory.open(Reader)
Please note that the last two methods must read the entire stream into memory and are not suitable for large streams.
All factory methods allow null and return an object representing an empty log file in this case.