tgroup identifies groups of rows in a table based on the
values in a given column or columns, and calculates statistical quantities
or otherwise collapses down the multiple values from other columns into
single values representing each group. It does the same job as a SELECT
... GROUP BY statement with aggregate functions in ADQL/SQL.
The keys parameter defines how input rows are grouped, and
the aggcols parameter defines what quantities to aggregate from the
rows in each group. keys specifies one or more values (column names
or expresssions) that must be the same for rows grouped together, while
aggcols specifies zero or more columns to be added based on the
content of rows in each group. The output table therefore contains one
column for each entry in keys and one column for each entry in
aggcols, and has one row for each group identified.
This command can therefore be used to count rows or calculate
statistical quantities per group. A number of statistical aggregation
methods are provided such as mean, median, minimum, maximum etc. For more
specialised requirements, for instance quantiles or custom statistics, you
can also use the array aggregators which generate an array containing all of
the values in the group, and operate on the resulting column using one of
the functions in the Arrays class.
By way of comparison, the tgroup invocation: stilts tgroup
in=t keys="year detector" aggcols="0;count;num
gmag;min;min_gmag gmag;mean" corresponds roughly to the ADQL query:
SELECT COUNT(*) AS num, MIN(gmag) AS min_gmag, MEAN(gmag), FROM t GROUP BY
year, detector
See also the tgridmap and tskymap commands, which
provide similar functionality where the grouping is over evenly spaced
numeric/coordinate values.
- ifmt=<in-format>
Specifies the format of the input table as specified by
parameter
in. The known formats are listed in SUN/256. This flag can be
used if you know what format your table is in. If it has the special value
(auto) (the default), then an attempt will be made to detect the format
of the table automatically. This cannot always be done correctly however, in
which case the program will exit with an error explaining which formats were
attempted. This parameter is ignored for scheme-specified tables.
- istream=true|false
If set true, the input table specified by the
in
parameter will be read as a stream. It is necessary to give the
ifmt
parameter in this case. Depending on the required operations and processing
mode, this may cause the read to fail (sometimes it is necessary to read the
table more than once). It is not normally necessary to set this flag; in most
cases the data will be streamed automatically if that is the best thing to do.
However it can sometimes result in less resource usage when processing large
files in certain formats (such as VOTable). This parameter is ignored for
scheme-specified tables.
- in=<table>
The location of the input table. This may take one of the
following forms:
- A filename.
- A URL.
- The special value "-", meaning standard input. In this
case the input format must be given explicitly using the ifmt
parameter. Note that not all formats can be streamed in this way.
- A scheme specification of the form
:<scheme-name>:<scheme-args>.
- A system command line with either a "<" character at
the start, or a "|" character at the end
("<syscmd" or "syscmd|"). This
executes the given pipeline and reads from its standard output. This will
probably only work on unix-like systems.
In any case, compressed data in one of the supported compression formats (gzip,
Unix compress or bzip2) will be decompressed transparently.
- icmd=<cmds>
Specifies processing to be performed on the input table
as specified by parameter
in, before any other processing has taken
place. The value of this parameter is one or more of the filter commands
described in SUN/256. If more than one is given, they must be separated by
semicolon characters (";"). This parameter can be repeated multiple
times on the same command line to build up a list of processing steps. The
sequence of commands given in this way defines the processing pipeline which
is performed on the table.
Commands may alteratively be supplied in an external file, by
using the indirection character '@'. Thus a value of
"@filename" causes the file filename to be read for
a list of filter commands to execute. The commands in the file may be
separated by newline characters and/or semicolons, and lines which are blank
or which start with a '#' character are ignored.
- ocmd=<cmds>
Specifies processing to be performed on the output table,
after all other processing has taken place. The value of this parameter is one
or more of the filter commands described in SUN/256. If more than one is
given, they must be separated by semicolon characters (";"). This
parameter can be repeated multiple times on the same command line to build up
a list of processing steps. The sequence of commands given in this way defines
the processing pipeline which is performed on the table.
Commands may alteratively be supplied in an external file, by
using the indirection character '@'. Thus a value of
"@filename" causes the file filename to be read for
a list of filter commands to execute. The commands in the file may be
separated by newline characters and/or semicolons, and lines which are blank
or which start with a '#' character are ignored.
- omode=out|meta|stats|count|checksum|cgi|discard|topcat|samp|tosql|gui
The mode in which the result table will be output. The
default mode is
out, which means that the result will be written as a
new table to disk or elsewhere, as determined by the
out and
ofmt parameters. However, there are other possibilities, which
correspond to uses to which a table can be put other than outputting it, such
as displaying metadata, calculating statistics, or populating a table in an
SQL database. For some values of this parameter, additional parameters
(
<mode-args>) are required to determine the exact behaviour.
Possible values are
- out
- meta
- stats
- count
- checksum
- cgi
- discard
- topcat
- samp
- tosql
- gui
Use the
help=omode flag or see SUN/256 for more information.
- out=<out-table>
The location of the output table. This is usually a
filename to write to. If it is equal to the special value "-" (the
default) the output table will be written to standard output.
This parameter must only be given if omode has its default
value of "out".
- ofmt=<out-format>
Specifies the format in which the output table will be
written (one of the ones in SUN/256 - matching is case-insensitive and you can
use just the first few letters). If it has the special value
"
(auto)" (the default), then the output filename will be
examined to try to guess what sort of file is required usually by looking at
the extension. If it's not obvious from the filename what output format is
intended, an error will result.
This parameter must only be given if omode has its default
value of "out".
- keys=<expr>
...
List of one or more space-separated words defining the
groups within which aggregation should be done. Each word can be a column name
or an expression using the expression language. Each expression will appear as
one of the columns in the output table. This list corresponds to the contents
of an ADQL/SQL
GROUP BY clause.
- aggcols=<expr>;<aggregator>[;<name>]
...
Defines the aggregate quantities to be calculated for
each group of input rows. Each quantity is defined by one entry in this list;
entries are space-separated, or can be given by multiple instances of this
parameter on the command line.
Each entry is composed of two or three tokens, separated by
semicolon (";") characters:
- <expr>: (required) column name, or expression using the
expression language, for the quantity to be aggregated
- <aggregator>: (required) aggregation method
- <name>: (optional) name of output column; if omitted, a name
based on the <expr> value will be used
The available <aggregator> values are as follows:
- count: counts the number of rows
- ngood: counts the number of non-blank items
- sum: the sum of all the combined values per bin
- mean: the mean of the combined values
- median: the median
- stdev: the sample standard deviation of the combined values
- stdev-pop: the population standard deviation of the combined
values
- max: records the maximum value
- min: records the minimum value
- array: collects all non-blank values into an array
- array-withblanks: collects all values into an array; blank values
are represented as zero for integers
- count-long: counts the number of rows, works for >2 billion
- ngood-long: counts the number of non-blank items, works for >2
billion
- runner=sequential|parallel|parallel<n>|partest
Selects the threading implementation, i.e. to what extent
processing is done in parallel. The options are currently:
- sequential: runs using only a single thread
- parallel: runs using multiple threads for large tables, with
parallelism given by the number of available processors
- parallel<n>: runs using multiple threads for large tables,
with parallelism given by the supplied value <n>
- partest: runs using multiple threads even when tables are small
(only intended for testing purposes)
Using parallel processing can speed up execution considerably;
however, depending on the I/O operations required, it can also slow it down
by disrupting patterns of disk access. If the content of a file is on a
solid state disk, or is already in cache for instance because a similar
command has been run recently, then parallel will probably be faster.
However, if the data is being read directly from a spinning disk, for
instance because the file is too large to fit in RAM, then sequential
or parallel<n> with a small <n> may be faster.
The value of this parameter should make only very tiny differences
to the output table. If you notice significant discrepancies please
report them.
- sort=true|false
Determines whether an attempt is made to sort the output
table by the values of the
keys expressions. This may not be possible
if no sort order is defined on the keys.
In most cases such sorting will be a small overhead on the rest of
the work done by this task, so the default is true but if ordering by
key is not useful you may save some resources by setting it false. If
no sorting is done, the output row order is undefined.
- cache=true|false
Determines whether the results of the aggregation
operation will be cached in random-access storage before output. This is set
true by default, since accessing rows of the calculated table may be somewhat
expensive, and most uses of the results will need all of the cells. But if you
anticipate making only a small number of accesses to the output table cells,
it could be more efficient to set this false.