csvtools

CsvTools provides transformation tools for CSV files. Supported transformation are the followings:

• projection (column filtering)

• selection (row filtering by row condition)

• sorting (rows ordered by given column(s))

• aggregation (applying given column function to given columns)

• extension (appending new column(s) to the CSV as result of given row functions)

• join of two CSV files

  • natural join
  • equijoin
  • outer left join
  • cross (full) join

Furthermore, CsvTools can be used as C++ library (libcsvtools).

Examples

The input is given as a well-formed CSV file (cells separated by comma (,), records separated by newline (\n) and the CSV file is also terminated by newline, first line is the CSV header, cells that need to be escaped are escaped by doublequote ("), each line contains as many cells (fields) as the header):

persons.csv

"ID","YoB","Name","Address"
0,1685,"Johann Sebastian Bach","Leipzig"
1,1756,"Wolfgang Amadeus Mozart","Salzburg"
2,1770,"Ludwig van Beethoven",Vienna
3,1810,"Frédéric Chopin",Paris
4,1984,"John ""Johnny"" Smith","London,
Soho"
5,1984,"Peter Carpenter","London"

Projection

For projection use csvproj. csvproj takes option -c <expr>, where expr defines columns or column interval.

• -c 0,1,2 denotes first (0), second (1) and third (2) columns, respectively. Note that column numbering begins with 0. This column sequence definition is equivalent with -c 0 -c 1 -c 2.
• -c 2,0 denotes third (2) and first (0) columns, respectively.
• -c 1,1 denotes the second (1) column twice.
• -c 0-2 denotes column interval [0,2], i.e., columns {0,1,2}.
• -c -3 denotes columns up to interval up to column 3, i.e., columns {0,1,2,3}.
• -c 3- denotes colums 3 and followers, i.e., columns {3,4,...,last}.
• -c 1,0-2 denotes columns {1,0,1,2}.
• -c r0 denotes the last column.
• -c r1 denotes the last but one column.
• -c r2- denotes the last three columns.
• -c -r1 denotes all columns except for the last one.
• Without any column interval option csvproj assumes that colums {} have to be projected, hence it produces an empty line for each input record.

cat persons.csv | csvproj -c 2 produces

"Name"
"Johann Sebastian Bach"
"Wolfgang Amadeus Mozart"
"Ludwig van Beethoven"
"Frédéric Chopin"
"John ""Johnny"" Smith"
"Peter Carpenter"

cat persons.csv | csvproj -c 3-,-0 produces

"Address","ID"
"Leipzig",0
"Salzburg",1
Vienna,2
Paris,3
"London,
Soho",4
"London",5

Selection

Use csvfilter for selecting rows by condition. The special option interpreted by csvfilter is -r <cond>, where cond defines the fitering condition. The condition is given as a relation between a column (or sequence of columns) and a string constant. The syntax of the condition is cond ::= <relation>:<expr>:str where relation is one of {=, !=, <, <=, >, >=}, expr is a column expression (see Projection) and str is a string constant. If multiple -r <cond> conditions are given, row must fulfill each of them to get selected.

• -r =:0:1 select rows where the first cell equals 1.

• -r =:1,3:foo,bar selects rows where columns 1 and 3 (separated by default output field separator, ,) give foo,bar. Equivalent with

  • -r =:3,1:bar,foo
  • -r =:1:foo -r =:3:bar

• -r >=:0,50 -r <:0:100 selects rows where the first column is at least 50 but less than 100.

cat persons.csv | csvfilter -r "<:1:1800" produces

"ID","YoB","Name","Address"
0,1685,"Johann Sebastian Bach","Leipzig"
1,1756,"Wolfgang Amadeus Mozart","Salzburg"
2,1770,"Ludwig van Beethoven",Vienna

Sorting

With csvsort the rows of CSV file can be sorted. Features:

• multi-column sort,
• ascending / descending sort,
• numeric / lexicographic sort,
• special handling of empty cells.

The order of the rows can by contolled with the -k {num}[d][n][e] option where num is a column number, d is flag for descending order, n flag is for numeric sort and e flag denotes that empty cells are greater than not empty cells. If multiple order columns are defined, the second, third, etc. definitions are taken into account only if the rows cannot be sorted with the first definition.

Examples:

• -k 0 sorts the rows ordered by the first column, lexicographically.
• -k 0 -k 1n sorts the rows ordered by the first column lexicographically, and if rows have the same value in the first column, their order is defined by the second column with numeric order.
• -k 0d rows ordered by first column, descending order.
• -k 0e rows ordered by first column, but rows with empty first cell go to the end.

Aggregation

With csvaggr a column (or a set of columns) can be aggregated. csvaggr supports the following column functions:

• sum
• min
• max
• count
• concat
• first
• last

Use csvaggr with option -a <name> <fun> <expr> where <name> will be the name of the result column, <fun> is the applied column function and <expr> is the column expression (see Projection). There can be given multiple -a options. If a column is not part of any column expression, it will be used as grouping column. Thus, every column either will be aggregated or will be a grouping column.

For csvaggr to work properly, it is requsite that the input csv data is sorted by the grouping columns.

Unnecessary columns cannot be dropped with csvaggr. Use csvproj to get rid of unused columns.

Summarized, for aggregating an arbitrary csv file, invoke the csvproj | csvsort | csvaggr process pipeline.

Examples:

• csvaggr -a cnt count - counts the records.
• csvaggr -a cnt count 1- counts the records grouped by the first column.
• csvaggr -a cnt count 1,3- counts the the records grouped by the first and the third columns.
• csvproj -c -1 | csvaggr -a sum sum 1 calculates the sum of values in the second column grouped by the first column.
• csvproj -c 4 | csvaggr -a min min 0 finds the minimum value of column #4.

Extension

With csvext new columns can be appended to the CSV data. The content of the cells in the new column is the result of a predefined function. The supported functions are

• int functions: add, sub, mul, div, mod, and, or
• float/numeric functions: int, fadd, fsub, fmul, fdiv, prec
• string functions: id, concat, length, index, substr, lower, upper
• branching functions: ifeq, iflt_str, iflt_int, ifle_str, ifle_int

The arguments of the functions can be either existing cells in the current row or constant values.

Use csvext with option -a <fun>:<name>, where <fun> is one of the predefined function, and is the name of the new, appended column. csvext accepts option -a multiple times, and the specified new columns will appear in the output in the same order as the were given at the command line.

The arguments of the function must be given by means of -pc <fpos>:<apos>:<cidx> (referring to an existing cell in column <cidx>) or -pv <fpos>:<apos>:<val> (for passing the constant value <val>). <fpos> is the function position, the index of the function (starting with 0), <apos> is the argument position within the function.

Examples:

• csvext -a id:dat -pv 0:0:missing appends a column with name "dat", and the cells content is "missing".
• csvext -a add:sum -pc 0:0:0 -pc 0:1:3 appends column "sum" with the sum of the 0th and 3rd columns.
• csvext -a add:inc10 -pc 0:0:2 -pv 0:1:10 appends column "inc10" with the content of the 2nd column increased by 10.

Joining CSV files

The command csvjoin supports joining of two csv files (left and right tables). The left table is read from stdin, whereas the right table must be defined by command line parameter -jf <filename>. The join type is also a mandatory parameter: -jt <join_type>. The following join_type values are available:

• natural recognizes common colums, and performs inner join based on the common columns.
• cross performs full join. Results in Descartes-product of the two input tables.
• inner for inner join.
• outer for outer left join.

In case of inner and outer join, the join columns of the left and the right tables must be provided: -jc <expr>:<expr>, where expr is a column expression (see Projection).

Dealing with Other-than-Comma Separated Values

CSV files are not always comma-separated. Some "CSV" files contain TAB- or semicolon-separated values. With csvconv these files can be converted into comma-separated files and CSV files can be converted into Any-character Separated Values file. Available options:

• -ifs <char> sets input field separator character
• -ofs <char> sets output field separator character
• -irs <char> sets input record separator character
• -ors <char> sets output record separator character
• -esc <strategy> sets global output cell escaping strategy
• -cesc <expr> <strategy> sets output cell escaping strategy for a set of columns
• -hesc <strategy> sets output cell escaping strategy for the header record

<expr> is a column expression (see Projection). The escaping strategy is one of the followings:

• all: every field will be escaped.
• preserve: already escaped fields remain escaped and escaping is introduced where necessary.
• resolve: same as preserve, except for the fields where field escaping was necessary at the input, but is not required at the output -- there fields will not be escaped.
• remove: remove escaping from fields if possible.

The default escaping strategy is preserve. cesc overrides esc, hesc overrides both esc and cesc.