README
¶
csvgen
csvgen is a simple code generator for CSV readers that write records directly into structs and not just slices of strings.
Usage
type SomeRecord struct {
CreatedAt time.Time
Description string
}
//go:generate csvgen --output gen_somereader.go
type SomeReader struct {
reader *csv.Reader
}
func (r *SomeReader) parseDate(ctx context.Context, raw string, format string) (time.Time, error) {
return time.Parse(format, raw)
}
Next, place a csvgen.yml file within the package where you've put the
go:generate call.
processors:
- type: SomeReader
recordType: SomeRecord
encoding: Windows1252
comma: ";"
fields:
CreatedAt:
column: 0
decoder:
name: ".parseDate"
options:
- "02.01.2006"
Description:
column: 1
When you now run go:generate, csvgen will create a gen_somereader.go file
in the same directory as the source file which provides a constructor function
for the struct as well as a Read and ReadAll function similar to the ones
provided by encoding/csv.Reader.
Configuration file
All the configuration what field should be filled based on what column in the
CSV file is handled within a configuration file that you can specify when
calling csvgen. As you saw in the example above, you specify a list of
"processors" that map a record to a reader. For each such mapping you can
define a field mapping with each field having a column number (0-indexed) and a
decoder.
The decoder's name, if starting with a . is mapped to a method within the
processor type, otherwise a function within the module is used.
Why no struct tags?
One alterative for the field format would have been to add all of that to the target record type. The problem with this is that it moves reader knowledge into a pure data object. That being said, we might add support for this in the future.
Directories