tsv-cast-header

#!/usr/bin/env python3
"""
usage: tsv-cast-header <target.tsv> <source.tsv>

Casts a <source.tsv> into the header of <target.tsv>.

Fields are reordered, dropped, and added as necessary.  Added fields will have
blank values.

No output will be emitted if <source.tsv> has no rows.  <target.tsv> must have
at least a header line.

All conversion is performed in a memory efficient manner, and inputs do not
need to be seekable.

---

This program exists because both `tsv-append` (from tsv-utils) and `csvtk
concat` are by themselves unsuitable for this task.

`tsv-append` is header line aware, but not field aware: it assumes all inputs
have the exact same header line and does no re-ordering, adding, or dropping of
fields.  It will happily mismatch fields between inputs and produce data lines
with too few or too many fields.

`csvtk concat` is field aware and DTRT, but it buffers each input completely
into memory, making it a non-starter for large dataset sizes.
"""
import csv
from argparse import ArgumentParser, RawDescriptionHelpFormatter
from sys import stdin, stdout, stderr, exit


cli = ArgumentParser(
    description     = __doc__.strip().split("\n---\n", 1)[0].split("\n\n", 1)[1],
    epilog          = __doc__.strip().split("\n---\n", 1)[1],
    formatter_class = RawDescriptionHelpFormatter)

cli.add_argument("target", metavar = "<target.tsv>")
cli.add_argument("source", metavar = "<source.tsv>")

args = cli.parse_args()


# Read header line of <target.tsv>
with open(args.target, "r", encoding = "utf-8", newline = "") as target:
    lines = csv.reader(target, dialect = "excel-tab")
    try:
        header = next(lines)
    except StopIteration:
        print(f"{cli.prog}: error: {args.target!r} (the target) appears to empty; it must contain at least a header line", file = stderr)
        exit(1)


# Set up output for casting from one dict to another
output = csv.DictWriter(
    stdout,
    header,
    restval = "",
    extrasaction = "ignore",
    dialect = "excel-tab",
    lineterminator = "\n")


# Cast <source.tsv>
with open(args.source, "r", encoding = "utf-8", newline = "") as source:
    input = csv.DictReader(source, dialect = "excel-tab")

    for i, row in enumerate(input):
        if i == 0:
            if not set(input.fieldnames) & set(output.fieldnames):
                print(f"{cli.prog}: error: {args.target!r} (the target) and {args.source!r} (the source) share no fields; they must share at least one", file = stderr)
                exit(1)

            output.writeheader()

        output.writerow(row)