uk.ac.starlink.parquet

Class ParquetTableBuilder

java.lang.Object

uk.ac.starlink.table.formats.DocumentedTableBuilder

uk.ac.starlink.parquet.ParquetTableBuilder

All Implemented Interfaces:

Documented, DocumentedIOHandler, TableBuilder

public class ParquetTableBuilder
extends DocumentedTableBuilder

TableBuilder for parquet files.

Since:

25 Feb 2021

Author:

Mark Taylor

Constructor Summary

Constructors

Constructor and Description
ParquetTableBuilder() Constructor.

Method Summary

Modifier and Type	Method and Description
boolean	canImport(DataFlavor flavor) Indicates whether this builder is able to turn a resource of media type indicated by flavor into a table.
boolean	canStream() Returns false; parquet metadata is in the footer.
boolean	docIncludesExample() Indicates whether the serialization of some (short) example table should be added to the user documentation for this handler.
Boolean	getCacheCols() Returns policy for table construction.
String	getFormatName() Returns the name of the format which can be read by this handler.
int	getReadThreadCount() Returns the number of read threads to use when caching column data.
boolean	getTryUrl() Indicates whether an attempt is made to open parquet files from non-file URLs.
String	getXmlDescription() Returns user-directed documentation in XML format.
StarTable	makeStarTable(DataSource datsrc, boolean wantRandom, StoragePolicy storage) Constructs a StarTable based on a given DataSource.
void	setCacheCols(Boolean cacheCols) Determines policy for table construction.
void	setReadThreadCount(int nThread) Sets the number of read threads to use when caching column data.
void	setTryUrl(boolean tryUrl) Configures whether an attempt is made to open parquet files from non-file URLs.
void	streamStarTable(InputStream istrm, TableSink sink, String pos) Reads a table from an input stream and writes it a row at a time to a sink.

Methods inherited from class uk.ac.starlink.table.formats.DocumentedTableBuilder

getExtensions, looksLikeFile

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface uk.ac.starlink.table.formats.DocumentedIOHandler

matchesExtension, readText, toLink

Constructor Detail

ParquetTableBuilder

public ParquetTableBuilder()

Constructor.

Method Detail

getFormatName

public String getFormatName()

Description copied from interface: TableBuilder

Returns the name of the format which can be read by this handler. Matching against this string may be used by callers to identify or select this handler from a list.

Returns:

one-word description of this handler's format

canStream

public boolean canStream()

Returns false; parquet metadata is in the footer.

Specified by:

canStream in class DocumentedTableBuilder

Returns:

true iff this handler can read from streams

docIncludesExample

public boolean docIncludesExample()

Description copied from interface: DocumentedIOHandler

Indicates whether the serialization of some (short) example table should be added to the user documentation for this handler. Binary formats, or instances for which the Documented.getXmlDescription() method already includes some example output, should return false.

Returns:

true if the user documentation would benefit from the addition of an example serialization

getXmlDescription

public String getXmlDescription()

Description copied from interface: Documented

Returns user-directed documentation in XML format.

The output should be a sequence of one or more <P> elements, using XHTML-like XML. Since rendering may be done in a number of contexts however, use of the full range of XHTML elements is discouraged. Where possible, the content should stick to simple markup such as the elements P, A, UL, OL, LI, DL, DT, DD EM, STRONG, I, B, CODE, TT, PRE.

Returns:

XML description of this object

makeStarTable

public StarTable makeStarTable(DataSource datsrc,
                               boolean wantRandom,
                               StoragePolicy storage)
                        throws IOException

Description copied from interface: TableBuilder

Constructs a StarTable based on a given DataSource. If the source is not recognised or this builder does not know how to construct a table from it, then a TableFormatException should be thrown. If this builder thinks it should be able to handle the source but an error occurs during processing, an IOException can be thrown.

The wantRandom parameter is used to indicate whether, ideally, a random-access table should be returned. There is no requirement for the builder to honour this request, but if it knows how to make both random and non-random tables, it can use this flag to decide which to return.

Note: the presence of the wantRandom parameter is somewhat misleading. TableBuilder implementations usually should, and do, ignore it (it would be removed from the interface if it were not for backward compatibility issues). Regardless of the value of this parameter, implementations should return a random-access table only if it is easy for them to do so; in particular they should not use the supplied storagePolicy, or any other resource-expensive measure, to randomise a sequential table just because the wantRandom parameter is true.

Parameters:

datsrc - the DataSource containing the table resource

wantRandom - whether, preferentially, a random access table should be returned

storage - a StoragePolicy object which may be used to supply scratch storage if the builder needs it

Returns:

a StarTable made out of datsrc

Throws:

TableFormatException - if the table is not of a kind that can be handled by this handler

IOException - if an unexpected I/O error occurs during processing

streamStarTable

public void streamStarTable(InputStream istrm,
                            TableSink sink,
                            String pos)
                     throws TableFormatException

Description copied from interface: TableBuilder

Reads a table from an input stream and writes it a row at a time to a sink. Not all implementations will be able to do this; for instance, extracting the table from the data may be a two-pass process. Implementations which are unable to perform this function should throw a TableFormatException.

The input stream should be prepared for use prior to calling this method, so implementations should not in general attempt to decompress or buffer istrm.

Parameters:

istrm - input stream containing table data

sink - destination of the table

pos - position identifier describing the location of the table within the stream; see DataSource.getPosition() (may be null)

Throws:

TableFormatException - if the table can't be streamed or the data is malformed

canImport

public boolean canImport(DataFlavor flavor)

Description copied from interface: TableBuilder

Indicates whether this builder is able to turn a resource of media type indicated by flavor into a table. It should return true if it thinks that its TableBuilder.streamStarTable(java.io.InputStream, uk.ac.starlink.table.TableSink, java.lang.String) method stands a reasonable chance of successfully constructing a StarTable from a DataSource whose input stream is described by the DataFlavor flavor. It will typically make this determination based on the flavor's MIME type.

This method should only return true if the flavor looks like it is targeted at this builder; for instance a builder which uses a text-based format should return false for a flavor which indicates a MIME type of text/plain.

This method is used in supporting drag and drop functionality (see StarTableFactory.canImport(java.awt.datatransfer.DataFlavor[])).

Parameters:

flavor - the DataFlavor whose suitability as stream input is to be assessed

Returns:

true iff this builder reckons it stands a good chance of turning a stream of type flavor into a StarTable

setCacheCols

@ConfigMethod(property="cachecols",
              usage="true|false|null",
              example="true",
              doc="<p>Forces whether to read all the column data at table load\ntime.  If <code>true</code>, then when the table is loaded,\nall data is read by column into local scratch disk files,\nwhich is generally the fastest way to ingest all the data.\nIf <code>false</code>, the table rows are read as required,\nand possibly cached using the normal STIL mechanisms.\nIf <code>null</code> (the default), the decision is taken\nautomatically based on available information.\n</p>")
public void setCacheCols(Boolean cacheCols)

Determines policy for table construction. If true, the makeStarTable method returns a CachedParquetStarTable and if false a SequentialParquetStarTable. If null, the decision is made automatically on the basis of whether it looks like random access is required and file size etc.

Parameters:

cacheCols - column data read policy

getCacheCols

public Boolean getCacheCols()

Returns policy for table construction.

Returns:

true for caching, false for read as required, null for adaptive

setReadThreadCount

@ConfigMethod(property="nThread",
              usage="<int>",
              example="4",
              doc="<p>Sets the number of read threads used for concurrently\nreading table columns if the columns are cached at load time\n- see the <code>cachecols</code> option.\nIf the value is &lt;=0 (the default), a value is chosen\nbased on the number of apparently available processors.\n</p>")
public void setReadThreadCount(int nThread)

Sets the number of read threads to use when caching column data. This is the value passed to the CachedParquetStarTable constructor, and ignored when constructing a SequentialParquetStarTable.

Parameters:

nThread - read thread count, or <=0 for auto

getReadThreadCount

public int getReadThreadCount()

Returns the number of read threads to use when caching column data.

Returns:

read thread count, or <=0 for auto

setTryUrl

@ConfigMethod(property="tryUrl",
              doc="<p>Whether to attempt to open non-file URLs as parquet files.\nThis usually seems to fail with a cryptic error message,\nso it is not attempted by default, but it\'s possible that with\nsuitable library support on the classpath it might work,\nso this option exists to make the attempt.\n</p>",
              example="true")
public void setTryUrl(boolean tryUrl)

Configures whether an attempt is made to open parquet files from non-file URLs.

Parameters:

tryUrl - true to attempt opening non-file URLs

getTryUrl

public boolean getTryUrl()

Indicates whether an attempt is made to open parquet files from non-file URLs.

Returns:

true to attempt non-file URLs, false to just give up