qsv 16.1.0

static USAGE: &str = r#"
Transforms CSV data so that all records have the same length. The length is
the length of the longest record in the data (not counting trailing empty fields,
but at least 1). Records with smaller lengths are padded with empty fields.

This requires two complete scans of the CSV data: one for determining the
record size and one for the actual transform. Because of this, the input
given must be a file and not stdin.

Alternatively, if --length is set, then all records are forced to that length.
This requires a single pass and can be done with stdin.

Usage:
    qsv fixlengths [options] [<input>]
    qsv fixlengths --help

fixlengths options:
    -l, --length <arg>     Forcefully set the length of each record. If a
                           record is not the size given, then it is truncated
                           or expanded as appropriate.
    -r, --remove-empty     Remove empty columns.
    -i, --insert <pos>     If empty fields need to be inserted, insert them
                           at <pos>. If <pos> is zero, then it is inserted
                           at the end of each record. If <pos> is negative, it
                           is inserted from the END of each record going backwards.
                           If <pos> is positive, it is inserted from the BEGINNING
                           of each record going forward. [default: 0]
    --quote <arg>          The quote character to use. [default: "]
    --escape <arg>         The escape character to use. When not specified,
                           quotes are escaped by doubling them.

Common options:
    -h, --help             Display this message
    -o, --output <file>    Write output to <file> instead of stdout.
    -d, --delimiter <arg>  The field delimiter for reading CSV data.
                           Must be a single character. (default: ,)
    -q, --quiet            Don't print removed column information.
"#;

use std::cmp;

use serde::Deserialize;

use crate::{
    CliResult,
    config::{Config, Delimiter},
    util,
};

#[derive(Deserialize)]
struct Args {
    arg_input:         Option<String>,
    flag_length:       Option<usize>,
    flag_insert:       i16,
    flag_remove_empty: bool,
    flag_quiet:        bool,
    flag_quote:        Delimiter,
    flag_escape:       Option<Delimiter>,
    flag_output:       Option<String>,
    flag_delimiter:    Option<Delimiter>,
}

pub fn run(argv: &[&str]) -> CliResult<()> {
    let args: Args = util::get_args(USAGE, argv)?;
    let mut config = Config::new(args.arg_input.as_ref())
        .delimiter(args.flag_delimiter)
        .quote(args.flag_quote.as_byte())
        .no_headers(true)
        .flexible(true);

    if let Some(escape) = args.flag_escape {
        config = config.escape(Some(escape.as_byte())).double_quote(false);
    }

    if config.is_stdin() && args.flag_remove_empty {
        return fail_incorrectusage_clierror!(
            "<stdin> cannot be used with --remove-empty. Please specify a file path."
        );
    }

    // First determine if we need to identify existing empty columns
    let col_is_empty_vec_opt = if args.flag_remove_empty {
        // First pass: identify existing empty columns
        let mut rdr = config.reader()?;
        let mut record = csv::ByteRecord::new();
        let mut col_is_empty_vec: Vec<bool> = Vec::new();
        let mut first = true;

        while rdr.read_byte_record(&mut record)? {
            if first {
                col_is_empty_vec = vec![true; record.len()];
                first = false;
            }

            for (i, field) in record.iter().enumerate() {
                if !field.is_empty() {
                    col_is_empty_vec[i] = false;
                }
            }
        }

        // Count and report removed columns
        let empty_count = col_is_empty_vec.iter().filter(|&&x| x).count();
        if !args.flag_quiet && empty_count > 0 {
            eprintln!("Removed {empty_count} empty column(s)");
        }
        Some(col_is_empty_vec)
    } else {
        None
    };

    let length = if let Some(length) = args.flag_length {
        if length == 0 {
            return fail_incorrectusage_clierror!("Length must be greater than 0.");
        }
        length
    } else if args.flag_remove_empty {
        // When removing empty columns, we need to determine the maximum length
        // after filtering out the empty columns. This requires another pass through the data.
        let mut rdr = config.reader()?;
        let mut record = csv::ByteRecord::new();
        let mut maxlen = 0_usize;
        // Get reference to vector indicating which columns are empty
        let col_is_empty_vec = col_is_empty_vec_opt.as_ref().unwrap();

        while rdr.read_byte_record(&mut record)? {
            // For each record, count how many fields would remain after filtering
            // out the empty columns. We use get_unchecked since we know i is in bounds.
            let filtered_len = record
                .iter()
                .enumerate()
                .filter(|(i, _)| unsafe { !*col_is_empty_vec.get_unchecked(*i) })
                .count();
            // Keep track of maximum filtered length seen so far
            maxlen = cmp::max(maxlen, filtered_len);
        }
        maxlen
    } else {
        if config.is_stdin() {
            return fail_incorrectusage_clierror!(
                "<stdin> cannot be used with if --length is not set. Please specify a file path."
            );
        }
        let mut maxlen = 0_usize;
        let mut rdr = config.reader()?;
        let mut record = csv::ByteRecord::new();
        while rdr.read_byte_record(&mut record)? {
            let mut nonempty_count = 0;
            for (index, field) in record.iter().enumerate() {
                if index == 0 || !field.is_empty() {
                    nonempty_count = index + 1;
                }
            }
            maxlen = cmp::max(maxlen, nonempty_count);
        }
        maxlen
    };

    let mut rdr = config.reader()?;
    let mut wtr = Config::new(args.flag_output.as_ref()).writer()?;
    let mut record = csv::ByteRecord::new();
    let mut record_work = csv::ByteRecord::new();
    let mut filtered_record = csv::ByteRecord::new();
    let mut field_idx: i16;

    let flag_insert = args.flag_insert;

    let insert_pos = if flag_insert < 0 {
        length as i16 + flag_insert
    } else {
        flag_insert
    };

    let mut current_len: usize;
    while rdr.read_byte_record(&mut record)? {
        // First remove existing empty columns if needed
        if let Some(ref is_empty) = col_is_empty_vec_opt {
            filtered_record.clear();
            for (i, field) in record.iter().enumerate() {
                if i < is_empty.len() && !is_empty[i] {
                    filtered_record.push_field(field);
                }
            }
            record.clone_from(&filtered_record);
        }

        // Then handle length adjustments by comparing target length with current record length
        current_len = record.len();
        match length.cmp(&current_len) {
            std::cmp::Ordering::Greater => {
                // Record is too short - need to add empty fields
                if flag_insert == 0 {
                    // No insert position specified - append empty fields at end
                    record.extend((0..length - current_len).map(|_| b""));
                } else {
                    // Insert empty fields at specified position
                    record_work.clear();
                    field_idx = 1_i16;
                    for field in &record {
                        if field_idx == insert_pos {
                            // When we reach insert position, add all needed empty fields
                            record_work.extend((0..length - current_len).map(|_| b""));
                        }
                        record_work.push_field(field);
                        field_idx += 1;
                    }
                    if record_work.len() < length {
                        // If we never hit insert position (it was past end of record),
                        // append remaining empty fields at the end
                        record_work.extend((0..length - record_work.len()).map(|_| b""));
                    }
                    record.clone_from(&record_work);
                }
            },
            std::cmp::Ordering::Less => {
                // Record is too long - truncate to target length
                record.truncate(length);
            },
            std::cmp::Ordering::Equal => {
                // Record is already correct length - no changes needed
            },
        }
        wtr.write_byte_record(&record)?;
    }
    Ok(wtr.flush()?)
}