Join two or more fits binary tables on specified columns
funjoin [switches] <ifile1> <ifile2> ... <ifilen> <ofile>
-a cols # columns to activate in all files -a1 cols ... an cols # columns to activate in each file -b 'c1:bvl,c2:bv2' # blank values for common columns in all files -bn 'c1:bv1,c2:bv2' # blank values for columns in specific files -j col # column to join in all files -j1 col ... jn col # column to join in each file -m min # min matches to output a row -M max # max matches to output a row -s # add 'jfiles' status column -S col # add col as status column -t tol # tolerance for joining numeric cols [2 files only]
funjoin joins rows from two or more (up to 32) \s-1FITS\s0 Binary Table files, based on the values of specified join columns in each file. \s-1NB:\s0 the join columns must have an index file associated with it. These files are generated using the funindex program.
The first argument to the program specifies the first input \s-1FITS\s0 table or raw event file. If \*(L"stdin\*(R" is specified, data are read from the standard input. Subsequent arguments specify additional event files and tables to join. The last argument is the output \s-1FITS\s0 file.
\s-1NB:\s0 Do not use Funtools Bracket Notation to specify \s-1FITS\s0 extensions and row filters when running funjoin or you will get wrong results. Rows are accessed and joined using the index files directly, and this bypasses all filtering.
The join columns are specified using the \-j col switch (which specifies a column name to use for all files) or with \-j1 col1, \-j2 col2, ... \-jn coln switches (which specify a column name to use for each file). A join column must be specified for each file. If both \-j col and \-jn coln are specified for a given file, then the latter is used. Join columns must either be of type string or type numeric; it is illegal to mix numeric and string columns in a given join. For example, to join three files using the same key column for each file, use:
funjoin -j key in1.fits in2.fits in3.fits out.fits
A different key can be specified for the third file in this way:
funjoin -j key -j3 otherkey in1.fits in2.fits in3.fits out.fits
The \-a \*(L"cols\*(R" switch (and \-a1 \*(L"col1\*(R", \-a2 \*(L"cols2\*(R" counterparts) can be used to specify columns to activate (i.e. write to the output file) for each input file. By default, all columns are output.
If two or more columns from separate files have the same name, the second (and subsequent) columns are renamed to have an underscore and a numeric value appended.
The \-m min and \-M max switches specify the minimum and maximum number of joins required to write out a row. The default minimum is 0 joins (i.e. all rows are written out) and the default maximum is 63 (the maximum number of possible joins with a limit of 32 input files). For example, to write out only those rows in which exactly two files have columns that match (i.e. one join):
funjoin -j key -m 1 -M 1 in1.fits in2.fits in3.fits ... out.fits
A given row can have the requisite number of joins without all of the files being joined (e.g. three files are being joined but only two have a given join key value). In this case, all of the columns of the non-joined file are written out, by default, using blanks (zeros or NULLs). The \-b c1:bv1,c2:bv2 and \-b1 'c1:bv1,c2:bv2' \-b2 'c1:bv1,c2 - bv2' ... switches can be used to set the blank value for columns common to all files and/or columns in a specified file, respectively. Each blank value string contains a comma-separated list of column:blank_val specifiers. For floating point values (single or double), a case-insensitive string value of \*(L"nan\*(R" means that the \s-1IEEE\s0 NaN (not\-a\-number) should be used. Thus, for example:
funjoin -b "AKEY:???" -b1 "A:-1" -b3 "G:NaN,E:-1,F:-100" ...
means that a non-joined \s-1AKEY\s0 column in any file will contain the string \*(L"???\*(R", the non-joined A column of file 1 will contain a value of \-1, the non-joined G column of file 3 will contain \s-1IEEE\s0 NaNs, while the non-joined E and F columns of the same file will contain values \-1 and \-100, respectively. Of course, where common and specific blank values are specified for the same column, the specific blank value is used.
To distinguish which files are non-blank components of a given row, the \-s (status) switch can be used to add a bitmask column named \*(L"\s-1JFILES\s0\*(R" to the output file. In this column, a bit is set for each non-blank file composing the given row, with bit 0 corresponds to the first file, bit 1 to the second file, and so on. The file names themselves are stored in the \s-1FITS\s0 header as parameters named \s-1JFILE1\s0, \s-1JFILE2\s0, etc. The \-S col switch allows you to change the name of the status column from the default \*(L"\s-1JFILES\s0\*(R".
A join between rows is the Cartesian product of all rows in one file having a given join column value with all rows in a second file having the same value for its join column and so on. Thus, if file1 has 2 rows with join column value 100, file2 has 3 rows with the same value, and file3 has 4 rows, then the join results in 2*3*4=24 rows being output.
The join algorithm directly processes the index file associated with the join column of each file. The smallest value of all the current columns is selected as a base, and this value is used to join equal-valued columns in the other files. In this way, the index files are traversed exactly once.
The \-t tol switch specifies a tolerance value for numeric columns. At present, a tolerance value can join only two files at a time. (A completely different algorithm is required to join more than two files using a tolerance, somethng we might consider implementing in the future.)
The following example shows many of the features of funjoin. The input files t1.fits, t2.fits, and t3.fits contain the following columns:
[sh] fundisp t1.fits AKEY KEY A B ----------- ------ ------ ------ aaa 0 0 1 bbb 1 3 4 ccc 2 6 7 ddd 3 9 10 eee 4 12 13 fff 5 15 16 ggg 6 18 19 hhh 7 21 22
fundisp t2.fits
\s-1AKEY\s0 \s-1KEY\s0 C D
----------- ------ ------ ------
iii 8 24 25
ggg 6 18 19
eee 4 12 13
ccc 2 6 7
aaa 0 0 1
fundisp t3.fits
\s-1AKEY\s0 \s-1KEY\s0 E F G
------------ ------ -------- -------- -----------
ggg 6 18 19 100.10
jjj 9 27 28 200.20
aaa 0 0 1 300.30
ddd 3 9 10 400.40
Given these input files, the following funjoin command:
funjoin -s -a1 "-B" -a2 "-D" -a3 "-E" -b \ "AKEY:???" -b1 "AKEY:XXX,A:255" -b3 "G:NaN,E:-1,F:-100" \ -j key t1.fits t2.fits t3.fits foo.fits
will join the files on the \s-1KEY\s0 column, outputting all columns except B (in t1.fits), D (in t2.fits) and E (in t3.fits), and setting blank values for \s-1AKEY\s0 (globally, but overridden for t1.fits) and A (in file 1) and G, E, and F (in file 3). A \s-1JFILES\s0 column will be output to flag which files were used in each row:
AKEY KEY A AKEY_2 KEY_2 C AKEY_3 KEY_3 F G JFILES ------------ ------ ------ ------------ ------ ------ ------------ ------ -------- ----------- -------- aaa 0 0 aaa 0 0 aaa 0 1 300.30 7 bbb 1 3 ??? 0 0 ??? 0 -100 nan 1 ccc 2 6 ccc 2 6 ??? 0 -100 nan 3 ddd 3 9 ??? 0 0 ddd 3 10 400.40 5 eee 4 12 eee 4 12 ??? 0 -100 nan 3 fff 5 15 ??? 0 0 ??? 0 -100 nan 1 ggg 6 18 ggg 6 18 ggg 6 19 100.10 7 hhh 7 21 ??? 0 0 ??? 0 -100 nan 1 XXX 0 255 iii 8 24 ??? 0 -100 nan 2 XXX 0 255 ??? 0 0 jjj 9 28 200.20 4
See funtools(7) for a list of Funtools help pages