Cloudera Morphlines provides a set of frequently-used high-level transformation and I/O commands that can be combined in application specific ways. This chapter outlines the currently available commands.

• readClob
• readCSV
• readLine
• readMultiLine

• addCurrentTime
• addLocalHost
• addValues
• addValuesIfAbsent
• callParentPipe
• contains
• convertTimestamp
• decodeBase64
• dropRecord
• equals
• extractURIComponents
• extractURIQueryParameters
• findReplace
• generateUUID
• grok
• if
• java
• logTrace, logDebug, logInfo, logWarn, logError
• not
• pipe
• separateAttachments
• setValues
• split
• toString
• translate
• tryRules

• readAvroContainer
• readAvro
• extractAvroTree
• extractAvroPaths

• readJson
• extractJsonPaths

• readSequenceFile

• detectMimeType

• decompress
• unpack

• convertHTML
• xquery
• xslt

• solrLocator
• loadSolr
• generateSolrSequenceKey
• sanitizeUnknownSolrFields
• tokenizeText

• solrCell

Cloudera Morphlines provides a set of frequently-used high-level transformation and I/O commands that can be combined in application specific ways. This chapter outlines the currently available commands.

Note that perhaps the most important property of the Morphlines framework is how easy it is to add new transformations and I/O commands and integrate existing functionality and third party systems. If none of the existing commands match your use case, you can easily write your own command and plug it in. Simply implement the Java interface Command or subclass AbstractCommand and add the resulting Java class to the classpath. No registration or other administrative action is required. Here are links to two sample command implementations:

• toString
  
• readLine
  
  Indeed, none of the standard commands are special or intrinsically known per se. All commands are implemented like this, even including standard commands such as pipe, if, and tryRules. This means your custom commands can even replace any standard commands, if desired.

This module contains standard I/O commands for tasks such as acting on single-line records, multi-line records, CSV files, and for converting bytes to strings.

The readClob command converts bytes to strings. It emits one record for the entire input stream of the first attachment, interpreting the stream as a Character Large Object (CLOB). The line is put as a string into the message output field.

The command provides the following configuration options:

Example usage:

readClob {
  charset : UTF-8
}

The readCSV command extracts zero or more records from the input stream of the first attachment of the record, representing a Comma Separated Values (CSV) file.

For the format see this article.

Some CSV files contain a header line that contains embedded column names. This command does not support reading and using such embedded column names as output field names because this is considered unreliable for production systems. If the first line of the CSV file is a header line, you must set the ignoreFirstLine option to true. You must explicitly define the columns configuration parameter in order to name the output fields.

The command provides the following configuration options:

If the parameter quoteChar is a String containing a single character then a quoted field can span multiple lines in the input stream, for example as shown in the following example CSV input containing a single record with three columns:

column0,"Look, new hot tub under redwood tree!
All bubbly!",column2

The above example can be parsed by specifying a double-quote character for the parameter quoteChar, using backslash syntax per the JSON specification, as follows:

readCSV {
  ...
  quoteChar : "\""

If the parameter commentPrefix is a String containing a single character then lines starting with that character are ignored as comments. Example:

#This is a comment line. It is ignored.

Example usage for CSV (Comma Separated Values):

readCSV {
  separator : ","
  columns : [Age,"",Extras,Type]
  ignoreFirstLine : false
  quoteChar : ""
  commentPrefix : ""
  trim : true
  charset : UTF-8
}

Example usage for TSV (Tab Separated Values):

readCSV {
  separator : "\t"
  columns : [Age,"",Extras,Type]
  ignoreFirstLine : false
  quoteChar : ""
  commentPrefix : ""
  trim : true
  charset : UTF-8
}

Example usage for SSV (Space Separated Values):

readCSV {
  separator : " "
  columns : [Age,"",Extras,Type]
  ignoreFirstLine : false
  quoteChar : ""
  commentPrefix : ""
  trim : true
  charset : UTF-8
}

Example usage for Apache Hive (Values separated by non-printable CTRL-A character):

readCSV {
  separator : "\u0001"
  columns : [Age,"",Extras,Type]
  ignoreFirstLine : false
  quoteChar : ""
  commentPrefix : ""
  trim : false
  charset : UTF-8
}

The readLine command emits one record per line in the input stream of the first attachment. The line is put as a string into the message output field. Empty lines are ignored.

The command provides the following configuration options:

Example usage:

readLine {
  ignoreFirstLine : true
  commentPrefix : "#"
  charset : UTF-8
}

The readMultiLine command is a multiline log parser that collapse multiline messages into a single record. It supports regex, what, and negate configuration parameters similar to logstash. The line is put as a string into the message output field.

For example, this can be used to parse log4j with stack traces. Also see https://gist.github.com/smougenot/3182192 and http://logstash.net/docs/1.1.13/filters/multiline.

The command provides the following configuration options:

Example usage:

# parse log4j with stack traces
readMultiLine {
  regex : "(^.+Exception: .+)|(^\\s+at .+)|(^\\s+\\.\\.\\. \\d+ more)|(^\\s*Caused by:.+)"
  what : previous
  charset : UTF-8
}

# parse sessions; begin new record when we find a line that starts with "Started session"
readMultiLine {
  regex : "Started session.*"
  what : next
  charset : UTF-8
}

This module contains standard transformation commands, such as commands for flexible log file analysis, regular expression based pattern matching and extraction, operations on fields for assignment and comparison, operations on fields with list and set semantics, if-then-else conditionals, string and timestamp conversions, scripting support for dynamic java code, a small rules engine, logging, and metrics and counters.

The addCurrentTime command adds the result of System.currentTimeMillis() as a Long integer to a given output field. Typically, a convertTimestamp command is subsequently used to convert this timestamp to an application specific output format.

The command provides the following configuration options:

Example usage:

addCurrentTime {}

The addLocalHost command adds the name or IP of the local host to a given output field.

The command provides the following configuration options:

Example usage:

addLocalHost {
  field : my_host
  useIP : false
}

The addValues command takes a set of outputField : values pairs and performs the following steps: For each output field, adds the given values to the field. The command can fetch the values of a record field using a field expression, which is a string of the form @{fieldname}.

Example usage:

addValues {
  # add values "text/log" and "text/log2" to the source_type output field
  source_type : [text/log, text/log2]

  # add integer 123 to the pid field
  pid : [123]

  # add all values contained in the first_name field to the name field
  name : "@{first_name}"
}

The addValuesIfAbsent command is the same as the addValues command, except that a given value is only added to the output field if it is not already contained in the output field.

Example usage:

addValuesIfAbsent {
  # add values "text/log" and "text/log2" to the source_type output field
  # unless already present
  source_type : [text/log, text/log2]

  # add integer 123 to the pid field, unless already present
  pid : [123]

  # add all values contained in the first_name field to the name field
  # unless already present
  name : "@{first_name}"
}

The callParentPipe command routes records to the enclosing pipe object. Recall that a morphline is a pipe. Thus, unless a morphline contains nested pipes, the parent pipe of a given command is the morphline itself, meaning that the first command of the morphline is called with the given record. Thus, the callParentPipe command effectively implements recursion, which is useful for extracting data from container file formats in elegant and concise ways. For example, you could use this to extract data from tar.gz files. This command is typically used in combination with the commands detectMimeType, tryRules, decompress, unpack, and possibly solrCell.

Example usage:

callParentPipe {}

For a real world example, see the solrCell command.

The contains command succeeds if one of the field values of the given named field is equal to one of the the given values, and fails otherwise. Multiple fields can be named, in which case the results are ANDed.

Example usage:

# succeed if the _attachment_mimetype field contains a value "avro/binary"
# fail otherwise
contains { _attachment_mimetype : [avro/binary] }

# succeed if the tags field contains a value "version1" or "version2",
# fail otherwise
contains { tags : [version1, version2] }

The convertTimestamp command converts the timestamps in a given field from one of a set of input date formats (in an input timezone) to an output date format (in an output timezone), while respecting daylight savings time rules. The command provides reasonable defaults for common use cases.

The command provides the following configuration options:

Example usage:

# convert the timestamp field to "yyyy-MM-dd'T'HH:mm:ss.SSSZ"
# The input may match one of "yyyy-MM-dd'T'HH:mm:ss.SSS'Z'"
# or "yyyy-MM-dd'T'HH:mm:ss" or "yyyy-MM-dd".
convertTimestamp {
  field : timestamp
  inputFormats : ["yyyy-MM-dd'T'HH:mm:ss.SSS'Z'", "yyyy-MM-dd'T'HH:mm:ss", "yyyy-MM-dd"]
  inputTimezone : America/Los_Angeles
  outputFormat : "yyyy-MM-dd'T'HH:mm:ss.SSS'Z'"
  outputTimezone : UTC
}

The decodeBase64 command converts a Base64 encoded String to a byte[] per Section 6.8. "Base64 Content-Transfer-Encoding" of RFC 2045. The command converts each value in the given field and replaces it with the decoded value.

The command provides the following configuration options:

Example usage:

decodeBase64 { 
  field : screenshot_base64
}

The dropRecord command silently consumes records without ever emitting any record. This is much like piping to /dev/null.

Example usage:

dropRecord {}

The equals command succeeds if all field values of the given named fields are equal to the the given values and fails otherwise. Multiple fields can be named, in which case a logical AND is applied to the results.

Example usage:

# succeed if the _attachment_mimetype field contains the value "avro/binary"
# and nothing else, fail otherwise
equals { _attachment_mimetype : [avro/binary] }

# succeed if the tags field contains nothing but the values "version1"
# and "highPriority", in that order, fail otherwise
equals { tags : [version1, highPriority] }

The extractURIComponents command extracts subcomponents from the URIs contained in the given input field and adds them to output fields with the given prefix, namely scheme, authority, host, port, path, query, fragment, schemeSpecificPart, userInfo.

The command provides the following configuration options:

Example usage:

extractURIComponents {
  inputField : my_uri
  outputFieldPrefix : uri_component_
}

For example, given the input field myUri with the value http://userinfo@www.bar.com:8080/errors.log?foo=x&bar=y&foo=z#fragment the expected output is as follows:

The extractURIQueryParameters command extracts the query parameters with a given name from the URIs contained in the given input field and appends them to the given output field.

The command provides the following configuration options:

Example usage:

extractURIQueryParameters {
  parameter : foo
  inputField : myUri
  outputField : my_query_params
}

For example, given the input field myUri with the value http://userinfo@www.bar.com/errors.log?foo=x&bar=y&foo=z#fragment the expected output record is:

my_query_params:x
my_query_params:z

The findReplace command examines each string value in a given field and replaces each substring of the string value that matches the given string literal or grok pattern with the given replacement.

This command also supports grok dictionaries and regexes in the same way as the grok command.

The command provides the following configuration options:

Example usage with grok pattern:

findReplace { 
  field : message
  dictionaryFiles : [cdk-morphlines-core/src/test/resources/grok-dictionaries]                               
  pattern : """%{WORD:myGroup}"""
  #pattern : """(\b\w+\b)"""      
  isRegex : true
  replacement : "${myGroup}!"
  #replacement : "$1!"
  #replacement : ""
  replaceFirst : false
}

Input: "hello world"

Expected output: "hello! world!"

The generateUUID command sets a universally unique identifier on all records that are intercepted. An example UUID is b5755073-77a9-43c1-8fad-b7a586fc1b97, which represents a 128bit value.

The command provides the following configuration options:

Example usage:

generateUUID {
  field : my_id
}

The grok command uses regular expression pattern matching to extract structured fields from unstructured log data.

This is well suited for syslog logs, apache, and other webserver logs, mysql logs, and in general, any log format that is generally written for humans and not computer consumption.

A grok command can load zero or more dictionaries. A dictionary is a file or string that contains zero or more REGEXNAME to REGEX mappings, one per line, separated by space, for example:

INT (?:[+-]?(?:[0-9]+))
HOSTNAME \b(?:[0-9A-Za-z][0-9A-Za-z-]{0,62})(?:\.(?:[0-9A-Za-z][0-9A-Za-z-]{0,62}))*(\.?|\b)

For example, the regex named "INT" is associated with the following regex pattern:

[+-]?(?:[0-9]+)

and matches strings like "123", whereas the regex named "HOSTNAME" is associated with the following regex pattern:

\b(?:[0-9A-Za-z][0-9A-Za-z-]{0,62})(?:\.(?:[0-9A-Za-z][0-9A-Za-z-]{0,62}))*(\.?|\b)

and matches strings like "www.cloudera.com".

Morphlines ships with several standard grok dictionaries.

A grok command can contain zero or more grok expressions. Each grok expression refers to a record input field name and can contain zero or more grok patterns. The following is an example grok expression that refers to the input field named "message" and contains two grok patterns:

expressions : {
  message : """\s+%{INT:pid} %{HOSTNAME:my_name_servers}"""
}

The syntax for a grok pattern is

%{REGEX_NAME:GROUP_NAME}

for example

%{INT:pid}

or

%{HOSTNAME:my_name_servers}

The REGEXNAME is the name of a regex within a loaded dictionary.

The GROUPNAME is the name of an output field.

If all expressions of the grok command match the input record, then the command succeeds and the content of the named capturing group is added to this output field of the output record. Otherwise, the record remains unchanged and the grok command fails, causing backtracking of the command chain.

Note: The morphline configuration file is implemented using the HOCON format (Human Optimized Config Object Notation). HOCON is basically JSON slightly adjusted for the configuration file use case. HOCON syntax is defined at HOCON github page and as such, multi-line strings are similar to Python or Scala, using triple quotes. If the three-character sequence """ appears, then all Unicode characters until a closing """ sequence are used unmodified to create a string value.

In addition, the grok command supports the following parameters:

Example usage:

# Index syslog formatted files
#
# Example input line:
#
# <164>Feb  4 10:46:14 syslog sshd[607]: listening on 0.0.0.0 port 22.
#
# Expected output record fields:
#
# syslog_pri:164
# syslog_timestamp:Feb  4 10:46:14
# syslog_hostname:syslog
# syslog_program:sshd
# syslog_pid:607
# syslog_message:listening on 0.0.0.0 port 22.
#
grok {
  dictionaryFiles : [cdk-morphlines-core/src/test/resources/grok-dictionaries]
  expressions : {
    message : """<%{POSINT:syslog_pri}>%{SYSLOGTIMESTAMP:syslog_timestamp} %{SYSLOGHOST:syslog_hostname} %{DATA:syslog_program}(?:\[%{POSINT:syslog_pid}\])?: %{GREEDYDATA:syslog_message}"""

    #message2 : "(?<queue_field>.*)"
    #message4 : "%{NUMBER:queue_field}"
  }
}

More example usage:

# Split a line on one or more whitespace into substrings,
# and add the substrings to the "columns" output field.
#
# Example input line with tabs:
#
# "hello\t\tworld\tfoo"
#
# Expected output record fields:
#
# columns:hello
# columns:world
# columns:foo
#
grok {
  expressions : {
    message : """(?<columns>.+?)(\s+|\z)"""
  }
  findSubstrings : true
}

Even more example usage:

# Index a custom variant of syslog files where subfacility is optional
#
# Example input line:
#
# <179>Jun 10 04:42:51 www.foo.com Jun 10 2013 04:42:51 : %myproduct-3-mysubfacility-123456: Health probe failed
#
# Expected output record fields:
#
# my_message_code:%myproduct-3-mysubfacility-123456
# my_product:myproduct
# my_level:3
# my_subfacility:mysubfacility
# my_message_id:123456
# syslog_message:%myproduct-3-mysubfacility-123456: Health probe failed
#
grok {
  dictionaryFiles : [cdk-morphlines-core/src/test/resources/grok-dictionaries]
  dictionaryString : """
    MY_CUSTOM_TIMESTAMP %{MONTH} %{MONTHDAY} %{YEAR} %{TIME}
  """

  expressions : {
    message : """<%{POSINT}>%{SYSLOGTIMESTAMP} %{SYSLOGHOST} %{MY_CUSTOM_TIMESTAMP} : (?<syslog_message>(?<my_message_code>%%{\w+:my_product}-%{\w+:my_level}(-%{\w+:my_subfacility})?-%{\w+:my_message_id}): %{GREEDYDATA})"""
  }
}

Note: An easy way to test grok out is to use the online grok debugger from the logstash project. There is also a corresponding YouTube video.

The if command consists of a chain of zero or more conditions commands, as well as an optional chain of zero or or more commands that are processed if all conditions succeed ("then commands"), as well as an optional chain of zero or more commands that are processed if one of the conditions fails ("else commands").

If one of the commands in the then chain or else chain fails, then the entire if command fails and any remaining commands in the then or else branch are skipped.

The command provides the following configuration options:

Example usage:

if {
  conditions : [
    { contains { _attachment_mimetype : [avro/binary] } }
  ]
  then : [
    { logInfo { format : "processing then..." } }
  ]
  else : [
    { logInfo { format : "processing else..." } }
  ]
}

The java command compiles and executes the given Java code block, wrapped into a Java method with a Boolean return type and several parameters, along with a Java class definition that contains the given import statements.

The parameters passed to the code block are record, config, parent, child, context, logger and are of types Record, Config, Command, Command, MorphlineContext, Logger, respectively.

Compilation is done in main memory, meaning without writing to the filesystem.

The result is an object that can be executed (and reused) any number of times. This is a high performance implementation, using an optimized variant of https://scripting.dev.java.net/" (JSR 223 Java Scripting). Calling eval() just means calling Method.invoke(), and, as such, has the same minimal runtime cost. As a result of the low cost, this command can be called on the order of 100 million times per second per CPU core on industry standard hardware.

The command provides the following configuration options:

Example usage:

java {
  imports : "import java.util.*;"
  code: """
    List tags = record.get("tags");
    if (!tags.contains("hello")) {
      return false;
    }
    tags.add("world");
    return child.process(record);
        """
}

These commands log a message to slf4j at the given log level. The command can fetch the values of a record field using a field expression, which is a string of the form @{fieldname}. The special field expression @{} can be used to log the entire record.

Example usage:

# log the entire record at DEBUG level to SLF4J
logDebug { format : "my record: {}", args : ["@{}"] }

More example usage:

# log the timestamp field and the entire record at INFO level to SLF4J
logInfo {
  format : "timestamp: {}, record: {}"
  args : ["@{timestamp}", "@{}"]
}

To automatically print diagnostic information such as the content of records as they pass through the morphline commands, consider enabling TRACE log level, for example by adding the following line to your log4j.properties file:

log4j.logger.com.cloudera.cdk.morphline=TRACE

The not command consists of one nested command, the Boolean return value of which is inverted.

Example usage:

if {
  conditions : [
    {
      not {
        grok {
          ... some grok expressions go here
        }
      }
    }
  ]
  then : [
    { logDebug { format : "found no grok match: {}", args : ["@{}"] } }
    { dropRecord {} }
  ]
  else : [
    { logDebug { format : "found grok match: {}", args : ["@{}"] } }
  ]
}

The pipe command has an identifier and contains a chain of zero or more commands, through which records get piped. A command transforms the record into zero or more records. The output records of a command are passed to the next command in the chain. A command has a Boolean return code, indicating success or failure. If any command in the pipe fails (meaning that it returns false), the whole pipe fails (meaning that it returns false), which causes backtracking of the command chain.

Because a pipe is itself a command, a pipe can contain arbitrarily nested pipes. A morphline is a pipe. "Morphline" is simply another name for the pipe at the root of the command tree.

The command provides the following configuration options:

Example usage demonstrating a pipe with two commands, namely addValues and logDebug:

pipe {
  id : my_pipe

  # Import all commands in these java packages, subpackages and classes.
  # Other commands on the Java classpath are not visible to this morphline.
  importCommands : [
    "com.cloudera.**",   # package and all subpackages
    "org.apache.solr.**", # package and all subpackages
    "com.mycompany.mypackage.*", # package only
    "com.cloudera.cdk.morphline.stdlib.GrokBuilder" # fully qualified class
  ]

  commands : [
    { addValues { foo : bar }}
    { logDebug { format : "output record: {}", args : ["@{}"] } }
  ]
}

The separateAttachments command emits one output record for each attachment in the input record's list of attachments. The result is many records, each of which has at most one attachment.

Example usage:

separateAttachments {}

The setValues command is the same as the addValues command, except that it first removes all values from the given output field, and then it adds new values.

Example usage:

setValues {
  # assign values "text/log" and "text/log2" to source_type output field
  source_type : [text/log, text/log2]

  # assign the integer 123 to the pid field
  pid : [123]

  # assign all values contained in the first_name field to the name field
  name : "@{first_name}"
}

The split command divides strings into substrings, by recognizing a separator (a.k.a. "delimiter") which can be expressed as a single character, literal string, regular expression, or grok pattern. This class provides the functionality of Guava's Splitter class as a morphline command, plus it also supports grok dictionaries and regexes in the same way as the grok command.

The command provides the following configuration options:

Example usage with literal string:

split { 
  inputField : message
  outputField : output          
  separator : ","        
  isRegex : false      
  #separator : """\s*,\s*"""        
  #isRegex : true      
  addEmptyStrings : false
  trim : true          
}

Input:

message:"_a ,_b_ ,c__"

Expected output:

output:_a
output:_b_
output:c__

Example usage with grok pattern or normal regex:

split { 
  inputField : message
  outputField : output
  # dictionaryFiles : [cdk-morphlines-core/src/test/resources/grok-dictionaries] 
  dictionaryString : """COMMA_SURROUNDED_BY_WHITESPACE \s*,\s*"""          
  separator : """%{COMMA_SURROUNDED_BY_WHITESPACE}"""        
  # separator : """\s*,\s*"""        
  isRegex : true      
  addEmptyStrings : true
  trim : false          
}

The toString command converts the Java objects in a given field using the Object.toString() method to their string representation, and optionally also applies the String.trim() method.

The command provides the following configuration options:

Example usage:

toString { field : source_type }

The translate command examines each value in a given field and replaces it with the replacement value defined in a given dictionary aka lookup hash table.

The command provides the following configuration options:

Example usage to translate Syslog severity level numeric codes to string labels:

translate {
  field : level
  dictionary : {
     0 : Emergency
     1 : Alert
     2 : Critical
     3 : Error
     4 : Warning
     5 : Notice
     6 : Informational
     7 : Debug
  }
  fallback : Unknown # if no fallback is defined and no match is found then the command fails
}

Input: level:0

Expected output: level:Emergency

Input: level:999

Expected output: level:Unknown

The tryRules command consists of zero or more rules. A rule consists of zero or more commands.

The rules of a tryRules command are processed in topdown order. If one of the commands in a rule fails, the tryRules command stops processing this rule, backtracks and tries the next rule, and so on, until a rule is found that runs all its commands to completion without failure (the rule succeeds). If a rule succeeds, the remaining rules of the current tryRules command are skipped. If no rule succeeds the record remains unchanged, but a warning may be issued or an exception may be thrown.

Because a tryRules command is itself a command, a tryRules command can contain arbitrarily nested tryRules commands. By the same logic, a pipe command can contain arbitrarily nested tryRules commands and a tryRules command can contain arbitrarily nested pipe commands. This helps to implement complex functionality for advanced usage.

The command provides the following configuration options:

Example usage:

tryRules {
  catchExceptions : false
  throwExceptionIfAllRulesFailed : true
  rules : [
    # next rule of tryRules cmd:
    {
      commands : [
        { contains { _attachment_mimetype : [avro/binary] } }
        ... handle Avro data here
        { logDebug { format : "output record: {}", args : ["@{}"] } }
      ]
    }

    # next rule of tryRules cmd:
    {
      commands : [
        { contains { _attachment_mimetype : [text/csv] } }
        ... handle CSV data here
        { logDebug { format : "output record: {}", args : ["@{}"] } }
      ]
    }
  ]
}

This module contains morphline commands for reading, extracting, and transforming Avro files and Avro objects.

The readAvroContainer command parses an InputStream or byte array that contains Avro binary container file data. For each Avro datum, the command emits a morphline record containing the datum as an attachment in the field _attachment_body.

The Avro schema that was used to write the Avro data is retrieved from the Avro container. Optionally, the Avro schema that shall be used for reading can be supplied with a configuration option; otherwise it is assumed to be the same as the writer schema.

Note: Avro uses Schema Resolution if the two schemas are different, e.g. if the reader schema is a subset of the writer schema for the purpose of efficient column projection.

The input stream or byte array is read from the first attachment of the input record.

The command provides the following configuration options:

Example usage:

# Parse Avro container file and emit a record for each avro object
readAvroContainer {
  # Optionally, require the input to match one of these MIME types:
  # supportedMimeTypes : [avro/binary]

  # Optionally, use this Avro schema in JSON format inline for reading:
  # readerSchemaString : """<json can go here>"""

  # Optionally, use this Avro schema file in JSON format for reading:
  # readerSchemaFile : /path/to/syslog.avsc
}

The readAvro command is the same as the readAvroContainer command except that the Avro schema that was used to write the Avro data must be explicitly supplied to the readAvro command because it expects raw Avro data without an Avro container and hence without a builtin writer schema.

Optionally, the Avro schema that shall be used for reading can be supplied with a configuration option; otherwise it is assumed to be the same as the writer schema.

Note: Avro uses Schema Resolution if the two schemas are different, e.g. if the reader schema is a subset of the writer schema for the purpose of efficient column projection.

The command provides the following configuration options:

Example usage:

# Parse Avro and emit a record for each avro object
readAvro {
  # supportedMimeTypes : [avro/binary]
  # readerSchemaString : """<json can go here>"""
  # readerSchemaFile : test-documents/sample-statuses-20120906-141433-subschema.avsc
  # writerSchemaString : """<json can go here>"""
  writerSchemaFile : test-documents/sample-statuses-20120906-141433.avsc
}

The extractAvroTree command converts an attached Avro datum to a morphline record by recursively walking the Avro tree and extracting all data into a single morphline record, with fields named by their path in the Avro tree.

The Avro input object is expected to be contained in the field _attachment_body, and typically placed there by an upstream readAvroContainer or readAvro command.

This kind of mapping is useful for simple Avro schemas, but for more complex schemas, this approach may be overly simplistic and expensive.

The command provides the following configuration options:

Example usage:

extractAvroTree {
  outputFieldPrefix : ""
}

The extractAvroPaths command uses zero or more Avro path expressions to extract values from an Avro object.

The Avro input object is expected to be contained in the field _attachment_body, and typically placed there by an upstream readAvroContainer or readAvro command.

Each path expression consists of a record output field name (on the left side of the colon ':') as well as zero or more path steps (on the right hand side), each path step separated by a '/' slash, akin to a simple form of XPath. Avro arrays are traversed with the '[]' notation.

The result of a path expression is a list of objects, each of which is added to the given record output field.

The path language supports all Avro concepts, including such concepts as nested structures, records, arrays, maps, and unions. The path language supports a flatten option that collects the primitives in a subtree into a flat output list.

The command provides the following configuration options:

Example usage:

extractAvroPaths {
  flatten : true
  paths : {
    my_price : /price

    my_docId : /docId
    my_links : /links
    my_links_backward : "/links/backward"
    my_links_forward : "/links/forward"
    my_name_language_code : "/name[]/language[]/code"
    my_name_language_country : "/name[]/language[]/country"
    my_name : /name

    /mymapField/foo/label : /mapField/foo/label/
  }
}

This module contains morphline commands for reading, extracting, and transforming JSON files and JSON objects.

The readJson command parses an InputStream or byte array that contains JSON data. For each top level JSON object, the command emits a morphline record containing the top level object as an attachment in the field _attachment_body.

The input stream or byte array is read from the first attachment of the input record.

The command provides the following configuration options:

Example usage:

readJson {}

Example usage with conversion from JSON to java.util.Map objects:

readJson {
  outputClass : java.util.Map
}

The extractJsonPaths command uses zero or more JSON path expressions to extract values from a Jackson JSON object of outputClass com.fasterxml.jackson.databind.JsonNode.

The JSON input object is expected to be contained in the field _attachment_body, and typically placed there by an upstream readJson command with outputClass : com.fasterxml.jackson.databind.JsonNode.

Each path expression consists of a record output field name (on the left side of the colon ':') as well as zero or more path steps (on the right hand side), each path step separated by a '/' slash, akin to a simple form of XPath. JSON arrays are traversed with the '[]' notation.

The result of a path expression is a list of objects, each of which is added to the given record output field.

The path language supports all JSON concepts, including such concepts as nested objects, arrays, etc. The path language supports a flatten option that collects the primitives in a subtree into a flat output list.

The command provides the following configuration options:

Example usage:

extractJsonPaths {
  flatten : true
  paths : {
    my_price : /price

    my_docId : /docId
    my_links : /links
    my_links_backward : "/links/backward"
    my_links_forward : "/links/forward"
    my_name_language_code : "/name[]/language[]/code"
    my_name_language_country : "/name[]/language[]/country"
    my_name : /name    
  }
}

The readSequenceFile command extracts zero or more records from the input stream of the first attachment of the record, representing an Apache Hadoop SequenceFile.

For the format and documentation of SequenceFiles see SequenceFile.

The command automatically handles Record-Compressed and Block-Compressed SequenceFiles.

The command provides the following configuration options:

Example usage:

readSequenceFile {
  keyField : "key"
  valueField : "value"
}

This module contains morphline commands for autodetecting MIME types from binary data. Depends on tika-core.

The detectMimeType command uses Apache Tika to autodetect the MIME type of the first attachment from the binary data. The detected MIME type is assigned to the _attachment_mimetype field.

The command provides the following configuration options:

Example usage:

detectMimeType {
  includeDefaultMimeTypes : false
  #mimeTypesFiles : [src/test/resources/custom-mimetypes.xml]
  mimeTypesString :
    """
    <mime-info>
      <mime-type type="text/space-separated-values">
        <glob pattern="*.ssv"/>
      </mime-type>

      <mime-type type="avro/binary">
        <magic priority="50">
          <match value="0x4f626a01" type="string" offset="0"/>
        </magic>
        <glob pattern="*.avro"/>
      </mime-type>

      <mime-type type="mytwittertest/json+delimited+length">
        <magic priority="50">
          <match value="[0-9]+(\r)?\n\\{&quot;" type="regex" offset="0:16"/>
        </magic>
      </mime-type>
    </mime-info>
    """
}

This module contains morphline commands for decompressing and unpacking files. Depends on tika-core and commons-compress.

The decompress command decompresses the first attachment, and supports gzip and bzip2 format.

Example usage:

decompress {}

The unpack command unpacks the first attachment, and supports tar, zip, and jar format. The command emits one record per contained file.

Example usage:

unpack {}

This module contains morphline commands for reading, extracting and transforming XML and HTML with XPath, XQuery and XSLT.

The convertHTML command converts any HTML to XHTML, using the TagSoup Java library.

Instead of parsing well-formed or valid XML, this command parses HTML as it is found in the wild: poor, nasty and brutish, though quite often far from short. TagSoup (and hence this command) is designed for people who have to process this stuff using some semblance of a rational application design. By providing this converter, it allows standard XML tools to be applied to even the worst malformed HTML.

The command reads an InputStream or byte array from the first attachment (field _attachment_body) of the input record, parses it as HTML and replaces the field with UTF-8 encoded XHTML.

The command provides the following configuration options:

Example usage:

convertHTML {
  charset : UTF-8
}

The xquery command parses an InputStream that contains an XML document and runs the given W3C XQuery over the XML document, using the Saxon Java library. For each item in the query result sequence, the command emits a corresponding morphline record.

The command reads an InputStream or byte array from the first attachment (field _attachment_body) of the input record.

Per the W3C specs, every valid XPath (e.g. //tweets/tweet[@color='blue']) is also a valid XQuery. If you are comfortable with XPath you are already almost there.

An XQuery result sequence contains zero or more items such as element nodes, attribute nodes, text nodes, atomic values, etc. For each item in the query result sequence, the morphline command converts the item to a record and pipes that record to the next morphline command. For an attribute node the attribute's XPath string value is filled into the record field named after the attribute name. For an element node the attributes and children of the element are treated as follows: The XPath string value of the attribute or child is filled into the record field named after the child's name.

The command provides the following configuration options:

Each fragment provides the following configuration options:

Note: The morphline configuration file is implemented using the HOCON format (Human Optimized Config Object Notation). HOCON is basically JSON slightly adjusted for the configuration file use case. HOCON syntax is defined at HOCON github page and as such, multi-line strings are similar to Python or Scala, using triple quotes. If the three-character sequence """ appears, then all Unicode characters until a closing """ sequence are used unmodified to create a string value.

Example usage:

xquery {
  fragments : [
    {
      fragmentPath : "/"
      externalVariables : {
        myVariable : "hello world"
      }            
      queryString : """
        (: Example test xquery :)        
        
        declare variable $myVariable as xs:string external;
        
        for $tweet in /tweets/tweet
        return 
        <user>
          {$tweet/@id} 
          {$tweet/user/@screen_name} 
          <myStatusCounts>{string($tweet/user/@statuses_count)}</myStatusCounts>
          <text>{$tweet/text}</text>
          <greeting>{$myVariable}</greeting>
        </user>
      """
    }
  ]
}

Here is an example output record for the query above:

id:11111112
screen_name:fake_user1
myStatusCounts:11111
text:Come, see new hot tub under Redwood Tree!
greeting:hello world

More examples can be found in the unit tests.

For more background, see resources such as the XQuery Primer and XQuery FLOWR Tutorial and XQuery: A Guided Tour and Wikipedia.

The xslt command parses an InputStream that contains an XML document and runs the given W3C XSL Transform over the XML document, using the Saxon Java library. For each item in the query result sequence, the command emits a corresponding morphline record.

The command reads an InputStream or byte array from the first attachment (field _attachment_body) of the input record.

An XSLT result sequence contains zero or more items such as element nodes, attribute nodes, text nodes, atomic values, etc. For each item in the query result sequence, the morphline command converts the item to a record and pipes that record to the next morphline command. For an attribute node the attribute's XPath string value is filled into the record field named after the attribute name. For an element node the attributes and children of the element are treated as follows: The XPath string value of the attribute or child is filled into the record field named after the child's name.

The command provides the following configuration options:

Each fragment provides the following configuration options:

Note: The morphline configuration file is implemented using the HOCON format (Human Optimized Config Object Notation). HOCON is basically JSON slightly adjusted for the configuration file use case. HOCON syntax is defined at HOCON github page and as such, multi-line strings are similar to Python or Scala, using triple quotes. If the three-character sequence """ appears, then all Unicode characters until a closing """ sequence are used unmodified to create a string value.

Example usage:

xslt {
  fragments : [
    {
      fragmentPath : "/"
      parameters : {
        myVariable : "hello world"
      }            
      queryString : """
        <!-- Example XSLT identity transformation -->
        <xsl:stylesheet xmlns:xsl="http://www.w3.org/1999/XSL/Transform" version="1.0">
        
        <xsl:template match="@*|node()">
          <xsl:copy>
            <xsl:apply-templates select="@*|node()"/>
          </xsl:copy>
        </xsl:template>
        
        </xsl:stylesheet>
      """
    }
  ]
}

More examples can be found in the unit tests.

For more background, see resources such as the XSLT Tutorial and Wikipedia.

This module contains morphline commands for Solr that higher level modules such as cdk-morphlines-solr-cell, search-mr, and search-flume depend on for indexing.

A solrLocator is a set of configuration parameters that identify the location and schema of a Solr server or SolrCloud. Based on this information a morphline Solr command can fetch the Solr index schema and send data to Solr. A solrLocator is not actually a command but rather a common parameter of many morphline Solr commands, and thus described separately here.

Example usage:

solrLocator : {
  # Name of solr collection
  collection : collection1

  # ZooKeeper ensemble
  zkHost : "127.0.0.1:2181/solr"

  # Max number of documents to pass per RPC from morphline to Solr Server
  # batchSize : 1000
}

The loadSolr command loads a record into a Solr server or MapReduce Reducer.

The command provides the following configuration options:

Example usage:

loadSolr {
  solrLocator : {
    # Name of solr collection
    collection : collection1

    # ZooKeeper ensemble
    zkHost : "127.0.0.1:2181/solr"

    # Max number of docs to pass per RPC from morphline to Solr Server
    # batchSize : 1000
  }
}

The generateSolrSequenceKey command assigns a record unique key that is the concatenation of the given baseIdField record field, followed by a running count of the record number within the current session. The count is reset to zero whenever a startSession notification is received.

For example, assume a CSV file containing multiple records but no unique ids, and the base_id field is the filesystem path of the file. Now this command can be used to assign the following record values to Solr's unique key field: $path#0, $path#1, ... $path#N.

The name of the unique key field is fetched from Solr's schema.xml file, as directed by the solrLocator configuration parameter.

The command provides the following configuration options:

Example usage:

generateSolrSequenceKey {
  baseIdField: ignored_base_id
  solrLocator : ${SOLR_LOCATOR}
}

The sanitizeUnknownSolrFields command sanitizes record fields that are unknown to Solr schema.xml by either deleting them (renameToPrefix parameter is absent or a zero length string) or by moving them to a field prefixed with the given renameToPrefix (for example, to use typical dynamic Solr fields).

Recall that Solr throws an exception on any attempt to load a document that contains a field that is not specified in schema.xml.

The command provides the following configuration options:

Example usage:

sanitizeUnknownSolrFields {
  solrLocator : ${SOLR_LOCATOR}
}

The tokenizeText command uses the embedded Solr/Lucene Analyzer library to generate tokens from a text string, without sending data to a Solr server.

This is useful for prototyping and debugging Solr applications. It is also useful for standalone usage outside of Solr, e.g. for extracting text features from documents for use with recommendation systems, clustering and classification applications.

The command provides the following configuration options:

Example usage:

tokenizeText {
  inputField : message
  outputField : tokens
  solrFieldType : text_en
  solrLocator : {
    # Name of solr collection
    collection : collection1

    # ZooKeeper ensemble
    zkHost : "127.0.0.1:2181/solr"
    
    # solrHomeDir : "example/solr/collection1"    
  }
}

For example, given the input field message with the value Hello World!\nFoo@Bar.com #%()123 the expected output record is:

tokens:hello
tokens:world
tokens:foo
tokens:bar.com
tokens:123

This example assumes the Solr field type named "text_en" is defined in schema.xml as shown in the following snippet:

...
<fieldType name="text_en" class="solr.TextField" positionIncrementGap="100">
  <analyzer type="index">
    <tokenizer class="solr.StandardTokenizerFactory"/>
    <filter class="solr.StopFilterFactory"
       ignoreCase="true"
       words="lang/stopwords_en.txt"
       enablePositionIncrements="true"
    />
    <filter class="solr.LowerCaseFilterFactory"/>
    <filter class="solr.EnglishPossessiveFilterFactory"/>
    <filter class="solr.KeywordMarkerFilterFactory" protected="protwords.txt"/>
    <filter class="solr.PorterStemFilterFactory"/>
  </analyzer>
</fieldType>

This module contains morphline commands for using SolrCell with Tika parsers. This includes support for types including HTML, XML, PDF, Word, Excel, Images, Audio, and Video.

The solrCell command pipes the first attachment of a record into one of the given Tika parsers, then maps the Tika output back to a record using SolrCell.

The Tika parser is chosen from the configurable list of parsers, depending on the MIME type specified in the input record. Typically, this requires an upstream detectMimeType command.

The command provides the following configuration options:

Example usage:

solrCell {
  solrLocator : ${SOLR_LOCATOR}

  # extract some fields
  capture : [content, a, h1, h2]

  # rename exif_image_height field to text field
  # rename a field to anchor field
  # rename h1 field to heading1 field
  fmap : { exif_image_height : text, a : anchor, h1 : heading1 }

  # xpath : "/xhtml:html/xhtml:body/xhtml:div/descendant:node()"

  parsers : [ # one or more nested Tika parsers
    { parser : org.apache.tika.parser.jpeg.JpegParser }
  ]
}

Here is a complex morphline that demonstrates integrating multiple heterogenous input file formats via a tryRules command, including Avro and SolrCell, using auto detection of MIME types via detectMimeType command, recursion via the callParentPipe command for unwrapping container formats, and automatic UUID generation:

morphlines : [
  {
    id : morphline1
    importCommands : ["com.cloudera.**", "org.apache.solr.**"]

    commands : [
      {
        # emit one output record for each attachment in the input
        # record's list of attachments. The result is a list of
        # records, each of which has at most one attachment.
        separateAttachments {}
      }

      {
        # used for auto-detection if MIME type isn't explicitly supplied
        detectMimeType {
          includeDefaultMimeTypes : true
          mimeTypesFiles : [target/test-classes/custom-mimetypes.xml]
        }
      }

      {
        tryRules {
          throwExceptionIfAllRulesFailed : true
          rules : [
            # next rule of tryRules cmd:
            {
              commands : [
                { logDebug { format : "hello unpack" } }
                { unpack {} }
                { generateUUID {} }
                { callParentPipe {} }
              ]
            }

            # next rule of tryRules cmd:
            {
              commands : [
                { logDebug { format : "hello decompress" } }
                { decompress {} }
                { callParentPipe {} }
              ]
            }

            # next rule of tryRules cmd:
            {
              commands : [
                {
                  readAvroContainer {
                    supportedMimeTypes : [avro/binary]
                    # optional, avro json schema blurb for getSchema()
                    # readerSchemaString : "<json can go here>"
                    # readerSchemaFile : /path/to/syslog.avsc
                  }
                }

                { extractAvroTree {} }

                {
                  setValues {
                    id : "@{/id}"
                    user_screen_name : "@{/user_screen_name}"
                    text : "@{/text}"
                  }
                }

                {
                  sanitizeUnknownSolrFields {
                    solrLocator : ${SOLR_LOCATOR}
                  }
                }
              ]
            }

            # next rule of tryRules cmd:
            {
              commands : [
                {
                  readJsonTestTweets {
                    supportedMimeTypes : ["mytwittertest/json+delimited+length"]
                  }
                }

                {
                  sanitizeUnknownSolrFields {
                    solrLocator : ${SOLR_LOCATOR}
                  }
                }
              ]
            }

            # next rule of tryRules cmd:
            {
              commands : [
                { logDebug { format : "hello solrcell" } }
                {
                  # wrap SolrCell around an Tika parsers
                  solrCell {
                    solrLocator : ${SOLR_LOCATOR}

                    capture : [
                      # twitter feed schema
                      user_friends_count
                      user_location
                      user_description
                      user_statuses_count
                      user_followers_count
                      user_name
                      user_screen_name
                      created_at
                      text
                      retweet_count
                      retweeted
                      in_reply_to_user_id
                      source
                      in_reply_to_status_id
                      media_url_https
                      expanded_url
                     ]

                    # rename "content" field to "text" fields
                    fmap : { content : text, content-type : content_type }

                    lowernames : true

                    # Tika parsers to be registered:
                    parsers : [
                      { parser : org.apache.tika.parser.asm.ClassParser }
                      { parser : org.gagravarr.tika.FlacParser }
                      { parser : org.apache.tika.parser.audio.AudioParser }
                      { parser : org.apache.tika.parser.audio.MidiParser }
                      { parser : org.apache.tika.parser.crypto.Pkcs7Parser }
                      { parser : org.apache.tika.parser.dwg.DWGParser }
                      { parser : org.apache.tika.parser.epub.EpubParser }
                      { parser : org.apache.tika.parser.executable.ExecutableParser }
                      { parser : org.apache.tika.parser.feed.FeedParser }
                      { parser : org.apache.tika.parser.font.AdobeFontMetricParser }
                      { parser : org.apache.tika.parser.font.TrueTypeParser }
                      { parser : org.apache.tika.parser.xml.XMLParser }
                      { parser : org.apache.tika.parser.html.HtmlParser }
                      { parser : org.apache.tika.parser.image.ImageParser }
                      { parser : org.apache.tika.parser.image.PSDParser }
                      { parser : org.apache.tika.parser.image.TiffParser }
                      { parser : org.apache.tika.parser.iptc.IptcAnpaParser }
                      { parser : org.apache.tika.parser.iwork.IWorkPackageParser }
                      { parser : org.apache.tika.parser.jpeg.JpegParser }
                      { parser : org.apache.tika.parser.mail.RFC822Parser }
                      { parser : org.apache.tika.parser.mbox.MboxParser,
                          additionalSupportedMimeTypes : [message/x-emlx] }
                      { parser : org.apache.tika.parser.microsoft.OfficeParser }
                      { parser : org.apache.tika.parser.microsoft.TNEFParser }
                      { parser : org.apache.tika.parser.microsoft.ooxml.OOXMLParser }
                      { parser : org.apache.tika.parser.mp3.Mp3Parser }
                      { parser : org.apache.tika.parser.mp4.MP4Parser }
                      { parser : org.apache.tika.parser.hdf.HDFParser }
                      { parser : org.apache.tika.parser.netcdf.NetCDFParser }
                      { parser : org.apache.tika.parser.odf.OpenDocumentParser }
                      { parser : org.apache.tika.parser.pdf.PDFParser }
                      { parser : org.apache.tika.parser.pkg.CompressorParser }
                      { parser : org.apache.tika.parser.pkg.PackageParser }
                      { parser : org.apache.tika.parser.rtf.RTFParser }
                      { parser : org.apache.tika.parser.txt.TXTParser }
                      { parser : org.apache.tika.parser.video.FLVParser }
                      { parser : org.apache.tika.parser.xml.DcXMLParser }
                      { parser : org.apache.tika.parser.xml.FictionBookParser }
                      { parser : org.apache.tika.parser.chm.ChmParser }
                    ]
                  }
                }

                { generateUUID { field : ignored_base_id } }

                {
                  generateSolrSequenceKey {
                    baseIdField: ignored_base_id
                    solrLocator : ${SOLR_LOCATOR}
                  }
                }

              ]
            }
          ]
        }
      }

      {
        loadSolr {
          solrLocator : ${SOLR_LOCATOR}
        }
      }

      {
        logDebug {
          format : "My output record: {}"
          args : ["@{}"]
        }
      }

    ]
  }
]

Note: More information on SolrCell can be found here: http://wiki.apache.org/solr/ExtractingRequestHandler