FlatFileReader

Short Description

FlatFileReader reads data from flat files, such as CSV (comma-separated values) file and delimited, fixed-length, or mixed text files.
The component can read a single file as well as a collection of files placed on a local disk or remotely. Remote files are accessible via HTTP, HTTPS, FTP, or SFTP protocols. Using this component, ZIP and TAR archives of flat files can be read. Also reading data from an input port, or dictionary is supported.
FlatFileReader has an alias - UniversalDataReader.

Ports

Metadata

FlatFileReader does not propagate metadata.
This component has metadata templates available.
The optional logging port for incorrect records has to define the following metadata structure - the record contains exactly five fields (named arbitrarily) of given types in the following order:

Metadata on output port 0 can use Autofilling Functions.
The source_timestamp and source_size functions work only when reading from a file directly (if the file is an archive or it is stored at a remote location, timestamp will be empty and size will be 0).

FlatFileReader Attributes

[1] Either both or neither of these attributes must be specified.

Details

Parsed data records are sent to the first output port. The component has an optional output logging port for getting detailed information about incorrect records. Only if Data Policy is set to controlled and a proper Writer (Trash or FlatFileWriter) is connected to port 1, all incorrect records together with the information about the incorrect value, its location and the error message are sent out through this error port.

Trimming Data

1. Input strings are implicitly (i.e. the Trim strings attribute kept at the default value) processed before converting to a value according to the field data type as follows:
   • Whitespace is removed from both the start and the end in the case of boolean, date, decimal, integer, long, or number.
   • Input string is set to a field including a leading and trailing whitespace in the case of byte, cbyte, or string.
2. If the Trim strings attribute is set to true, all leading and trailing whitespace characters are removed. A field composed of only whitespaces is transformed to null (zero length string). The false value implies preserving all leading and trailing whitespace characters. Remember that input string representing a numerical data type or boolean can not be parsed including whitespace. Thus, use the false value carefully.
3. Both the Skip leading blanks and Skip trailing blanks attributes have a higher priority than Trim strings. So, the input strings trimming will be determined by the true or false values of these attributes, regardless the Trim strings value.

Data Parsers

1. org.jetel.data.parser.SimpleDataParser - is a very simple but fast parser with limited validation, error handling and functionality. The following attributes are not supported:
   • Trim strings
   • Skip leading blanks
   • Skip trailing blanks
   • Incremental reading
   • Number of skipped records
   • Max number of records
   • Quoted strings
   • Missing column handling
   • Treat multiple delimiters as one
   • Skip rows
   • Verbose

On top of that, you cannot use metadata containing at least one field with one of these attributes:

• the field is fixed-length
• the field has no delimiter or, on the other hand, more of them
• Shift is not null (see Details Pane)
• Autofilling set to true
• the field is byte-based

2. org.jetel.data.parser.DataParser - an all-round parser working with any reader settings
3. org.jetel.data.parser.CharByteDataParser - can be used whenever metadata contain byte-based fields mixed with char-based ones. A byte-based field is a field of one of these types: byte, cbyte or any other field whose format property starts with the "BINARY:" prefix. See Numeric Format > Binary Formats.
4. org.jetel.data.parser.FixLenByteDataParser - used for metadata with byte-based fields only. It parses sequences of records consisting of a fixed number of bytes.

Choosing org.jetel.data.parser.SimpleDataParser while using Quoted strings will cause the Quoted strings attribute to be ignored.

Tips & Tricks

• Handling records with large data fields:
  FlatFileReader can process input strings of even hundreds or thousands of characters when you adjust the field and record buffer sizes. Just increase the following properties according to your needs: Record.MAX_RECORD_SIZE for record serialization, DataParser.FIELD_BUFFER_LENGTH for parsing and DataFormatter.FIELD_BUFFER_LENGTH for formatting. Finally, don’t forget to increase the DEFAULT_INTERNAL_IO_BUFFER_SIZE variable to be at least 2* MAX_RECORD_SIZE. For information on how to change these property variables, see Engine Configuration.

Examples

Processing files with headers

If the first rows of your input file do not represent real data but field labels instead, set the Number of skipped records attribute. If a collection of input files with headers is read, set the Number of skipped records per source.

Handling typist’s error when creating the input file manually

If you wish to ignore accidental errors in delimiters (such as two semicolons instead of a single one as defined in metadata when the input file is typed manually), set the Treat multiple delimiters as one attribute to true. All redundant delimiter characters will be replaced by the proper one.

Incremental reading

Incremental Reading allows you to read only new records from a file. This can be done by setting the Incremental key and Incremental file attributes.

Let us have a list of customers in a file customers.dat. Each record in the file consists of a date, first name and last name:

2018-02-01 23:58:02|Rocky|Whitson
2018-02-01 23:59:56|Marisa|Callaghan
2018-03-01 00:03:12|Yaeko|Gonzale
2018-03-01 00:15:41|Jeana|Rabine
2018-03-01 00:32:22|Daniele|Hagey

Read the file, then add a new record and run the graph again reading only the new record.

Solution
In the output metadata, create the date, firstName and lastName fields. Set their data types to date, string and string, respectively.
Use the Incremental key and Incremental file attributes.

If the first row of the input file is not a header, remember to set the Number of skipped records attribute to 0.

After the first read, the output file contains five records.

Now, add a new record to the file, for example:
2018-03-01 00:51:31|Nathalie|Mangram
and run the graph again.

This time, only the new record is written to the output file, ignoring the previously processed records.

Best Practices

We recommend users to explicitly specify encoding of the input file (with the Charset attribute). It ensures better portability of the graph across systems with different default encoding.
The recommended encoding is UTF-8.

Troubleshooting

With default charset (UTF-8), FlatFileReader cannot parse csv files with binary data. To parse csv files with binary data, change Charset attribute.

See also

FlatFileWriter
UniversalDataReader
Common Properties of Components
Specific Attribute Types
Common Properties of Readers