FixedLengthUtil.qm

# -*- mode: qore; indent-tabs-mode: nil -*-
# @file FixedLengthUtil.qm Qore user module for working with files with fixed lines length

/*  FixedLengthUtil.qm Copyright 2012 - 2015 Qore Technologies, sro

    Permission is hereby granted, free of charge, to any person obtaining a
    copy of this software and associated documentation files (the "Software"),
    to deal in the Software without restriction, including without limitation
    the rights to use, copy, modify, merge, publish, distribute, sublicense,
    and/or sell copies of the Software, and to permit persons to whom the
    Software is furnished to do so, subject to the following conditions:

    The above copyright notice and this permission notice shall be included in
    all copies or substantial portions of the Software.

    THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
    IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
    FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
    AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
    LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
    FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
    DEALINGS IN THE SOFTWARE.
*/

# minimum required Qore version
%requires qore >= 0.8.12
%requires Util
%require-types
%enable-all-warnings
%new-style

module FixedLengthUtil {
    version = "1.0";
    desc    = "user module for working with files with fixed length lines";
    author  = "Jiri Vaclavik <jiri.vaclavik@qoretechnologies.com>, Zdenek Behan <zdenek.behan@qoretechnologies.com>";
    url     = "http://qore.org";
    license = "MIT";
}

/* see release notes below for version history
*/

/** @mainpage FixedLengthUtil Module

    @tableofcontents

    @section fixedlengthutilintro Introduction to the FixedLengthUtil Module

    The %FixedLengthUtil module provides functionality for parsing files with fixed length lines. This means that we have at least one line type and each line type is described as several data items with fixed length.

    To use this module, use \c "%requires FixedLengthUtil" in your code.

    All the public symbols in the module are defined in the FixedLengthUtil namespace.

    Currently the module provides the following classes:
    - @ref FixedLengthUtil::FixedLengthAbstractIterator "FixedLengthAbstractIterator": Iterator class allowing for fixed-length-line data to be processed line by line on a record basis
    - @ref FixedLengthUtil::FixedLengthFileIterator "FixedLengthFileIterator": Iterator class allowing for fixed-length-line file to be processed line by line on a record basis
    - @ref FixedLengthUtil::FixedLengthDataIterator "FixedLengthDataIterator": Iterator class allowing for fixed-length-line strings to be processed line by line on a record basis
    - @ref FixedLengthUtil::FixedLengthAbstractWriter "FixedLengthAbstractWriter": Fixed length line writer
    - @ref FixedLengthUtil::FixedLengthFileWriter "FixedLengthFileWriter": Fixed length line writer into a file
    - @ref FixedLengthUtil::FixedLengthDataWriter "FixedLengthDataWriter": Fixed length line writer into a variable

    @section fixedlengthglobals Global Options

    Valid options are:
    - \c "date_format": the default date format for \c "date" fields (see @ref date_formatting for the value in this case)
    - \c "encoding": the output encoding for strings parsed or returned
    - \c "file_flags": additional writer @ref file_open_constants; @ref Qore::O_WRONLY | @ref Qore::O_CREAT are used by default. Use eg. @ref Qore::O_EXCL to ensure not to overwrite the target or @ref Qore::O_TRUNC to replace any existing file
    - \c "ignore_empty": if @ref Qore::True "True" then ignore empty lines
    - \c "number_format": the default number format for \c "float" or \c "number" fields (see @ref Qore::parse_number() and @ref Qore::parse_float() for the value in these cases)
    - \c "timezone": a string giving a time zone region name or an integer offset in seconds east of UTC

    @section fixedlengthspec Specification Hash

    Fixed length specification hash is in the form of a hash, where each hash key is the name of a record, and each value
    is a @ref fixedlengthspecrecordhash "record description hash" describing the record; see the following example:
    @code
# the following spec is suitable for input and output
const Specs = (
    "header": (
        "flow_type": ("length": 3, "type": "string"),
        "record_type": ("length": 3, "type": "int", "padding_char": "0"),
        "number_of_records": ("length": 8, "type": "int", "padding_char": "0"),
    ),
    "line": (
        "flow_type": ("length": 3, "type": "string"),
        "record_type": ("length": 3, "type": "int", "padding_char": "0"),
        "processing_id": ("length": 10, "type": "int", "padding_char": "0"),
        "processing_name": ("length": 10, "type": "string"),
        "po_number": ("length": 10, "type": "int", "padding_char": "0"),
    ),
    "trailer": (
        "flow_type": ("length": 3, "type": "string"),
        "record_type": ("length": 3, "type": "int", "padding_char": "0"),
        "number_of_records": ("length": 8, "type": "int", "padding_char": "0"),
    ),
);
    @endcode

    In the example above, \c "header", \c "line", and \c "trailer" are record names, and the values of each key are @ref fixedlengthspecrecordhash "record description hashes".

    @subsection fixedlengthspecrecordhash Record Description Hash

    Each record will have a number of fields described in the record description hash.  The record description hash keys
    represent the names of the fields, and the values are @ref fixedlengthspecrow "field specification hashes".

    In the \c "header" record in the example above, the fields are \c "flow_type", \c "record_type",
    and \c "number_of_records", and the values of each of those keys are @ref fixedlengthspecrow "field specification hashes"
    for the given fields.

    @subsection fixedlengthspecrow Field Specification Hash

    The field specification hash has the following format:

    |!Key|!Type|!Description
    | \c length|integer|the size of the field in bytes
    | \c type|string|the type of data bound to the field @ref fixedlengthspectypes
    | \c format|string|a date mask if the \a type of the field is \c "date"; see @ref date_formatting for more information
    | \c timezone|string|override global timezone for current "date" field
    | \c padding|string|set padding of the field "left" (default) or "right"; used only in writers; if not given then the default padding depends on the field's \a type: \c "int" fields get left padding (right justification) and all others get right padding (left justification)
    | \c padding_char|string|a string with size 1 to use for \c padding. Default \c " " (space). Used only in writers

    @subsection fixedlengthspectypes Field Data Types

    - \c "date"
    - \c "float"
    - \c "int"
    - \c "number"
    - \c "string"

    @section fixedlengthformat Fixed Length Data Format

    Input and output data are formatted in a hash with two mandatory keys:
    - \c "type": a string with name of the type
    - \c "record": a hash with line data in field - value map

    @code
    ("type": "type1", "record": {"col1" : 11111, "col2" : "bb"}),
    @endcode

    <b>Example of reading:</b>
    @code
#!/usr/bin/env qr

%requires FixedLengthUtil

hash specs = (
    "type1" : (
        "col1": ("length": 5, "type": "int"),
        "col2": ("length": 2, "type": "string"),
    ),
    "type2": (
        "col3": ("length": 1, "type": "string"),
        "col4": ("length": 3, "type": "string"),
        "col5": ("length": 8, "type": "date", "format": "DDMMYYYY",
        # "timezone" : "Europe/Prague", # use global if omitted
        ),
    ),
);

hash global_options = (
    "encoding"     : "UTF-8",
    "eol"          : "\n",
    "ignore_empty" : True,
    "timezone"     : "Europe/Prague", # used if not overridden in a date field specification
);

FixedLengthFileIterator i(file, specs, global_options);
while (i.next()) {
    operation_with_hash(i.getValue())
}
    @endcode

    <b>Example of writing:</b>
    @code
#!/usr/bin/env qr

%requires FixedLengthUtil

list data = (
    ("type": "type1", "record": {"col1" : 11111, "col2" : "bb"}),
    ("type": "type2", "record": {"col3" : "c", "col4" : "ddd", "col5" : "31122014"}),
    ("type": "type1", "record": {"col1" : 22222, "col2" : "gg"}),
);

hash specs = (
    "type1" : (
        "col1": ("length": 5, "type": "int"),
        "col2": ("length": 2, "type": "string"),
    ),
    "type2" : (
        "col3" : ("length": 1, "type": "string"),
        "col4" : ("length" : 3, "type": "string"),
        "col5" : ("length" : 8, "type": "date", "format": "DDMMYYYY", "timezone" : "Europe/Prague"),
    ),
);

hash global_options = (
    "eol" : "\n",
);

FixedLengthFileWriter w(file, specs, global_options);
w.write(data);
    @endcode

    @section fixedlengthutil_relnotes Release Notes

    @subsection fixedlengthutil_v1_0 Version 1.0
    - initial version of module
*/

#! The FixedLengthUtil namespace contains all the definitions in the FixedLengthUtil module
public namespace FixedLengthUtil {
    #! Unix end of line character sequence (for new OS X too)
    public const EOL_UNIX = "\n";
    #! MS DOS/Windows end of line character sequence
    public const EOL_WIN = "\r\n";
    #! Old (pre-OSX) Macintosh end of line character sequence
    public const EOL_MACINTOSH = "\r";

    # helper list of end of line values
    const EOLS = (EOL_UNIX, EOL_WIN, EOL_MACINTOSH);
}

#! Structured line iterator for abstract data allowing efficient "pipelined" processing
public class FixedLengthUtil::FixedLengthAbstractIterator {
    private {
        hash m_specs;
        hash m_line_lengths;
        string m_state;
        #! hash of conflicting records; key = record length, value = list of conflicting record names
        hash m_conflicts;
        *hash m_opts; # encoding, eol, ignore_empty, timezone, etc
    }

    #! Instantiates the FixedLengthAbstractIterator object
    /**
        @param spec Fixed-length line specification; see @ref fixedlengthspec for more information
        @param opts Global options; see @ref fixedlengthglobals for more information

        @throw FIXED-LENGTH-UTIL-INVALID-SPEC invalid record specification; invalid type or missing field length
    */
    constructor(hash spec, *hash opts) {
        m_specs           = spec;
        m_opts            += checkOptions(opts);

        # Determine line lengths of each spec, for use by the default identifyType() method. We do this by building a reverse length -> type table.
        m_line_lengths = hash();
        foreach string k in (m_specs.keyIterator()) {
            if (m_specs{k}.typeCode() != NT_HASH)
                throw "FIXED-LENGTH-UTIL-INVALID-SPEC", sprintf("expecting a record description hash assigned to record key %y; got type %y instead (value: %y)", k, m_specs{k}.type(), m_specs{k});
            int len = 0;
            foreach string c in (m_specs{k}.keyIterator()) {
                if (!m_specs{k}{c}.length)
                    throw "FIXED-LENGTH-UTIL-INVALID-SPEC", sprintf("Length missing for field %y", c);
                len += m_specs{k}{c}.length;
            }
            # If any two type line lengths are not unique, save the conflicting length so that only conflicting records must be identified
            if (m_line_lengths{len})
                m_conflicts{len} += (k,);
            else
                m_line_lengths{len} = k;
        }
    }

    abstract *string getLine();

    #! Returns the current record as a hash
    /** @par Example:
        @code
my hash $h = $i.getValue();
        @endcode

        @return The current record as a hash with the following keys:
        - \c "type": a string giving the record type name
        - \c "record": a hash giving the parsed record data
    */
    *hash getValue() {
        *string line = getLine();

        if (!line)
            return;

        hash result = hash();

        string type = identifyType(line);
        result.type = type;
        if (!checkTransition(m_state, type)) {
            throw "FIXED-LENGTH-UTIL-INVALID-TRANSITION", sprintf("record %y cannot follow record %y for line %y", (type ?? "<START>"), (type ?? "<START>"), line);
        }
        m_state = type;
        int pos = 0;
        foreach string col in (m_specs{type}.keyIterator()) {
            hash field = m_specs{type}{col};
            if (!field.timezone) {
                field.timezone = m_opts.timezone;
            }
            result.record{col} = transform(line.substr(pos, field.length), field);
            pos += field.length;
        }
        return result;
    }

    any transform(any value, hash type) {
        switch (type.type) {
            case "int": {
                return int(value);
            }

            case NOTHING:
            case "string": {
                if (value.typeCode() == NT_STRING) {
                    trim value;
                    return m_opts.encoding ? convert_encoding(value, m_opts.encoding) : value;
                }
                return string(value);
            }

            case "float": {
                if (value.typeCode() == NT_STRING) {
                    *string nf = type.format ? type.format : m_opts.number_format;
                    return nf ? parse_float(value, nf) : float(value);
                }
                return float(value);
            }
            case "number": {
                if (value.typeCode() == NT_STRING) {
                    *string nf = type.format ? type.format : m_opts.number_format;
                    return nf ? parse_number(value, nf) : number(value);
                }
                return number(value);
            }
            case "date": {
                if (value.empty()) {
                    return 1970-01-01;
                }
                TimeZone tz;
                if (type.timezone)
                    tz = type.timezone.type() == "object" ? type.timezone : new TimeZone(type.timezone);
                else if (m_opts.timezone)
                    tz = m_opts.timezone;

                int t = value.typeCode();
                if (t == NT_STRING) {
                    *string fmt = type.format;
                    if (!fmt)
                        fmt = m_opts.date_format;

                    return tz ? tz.date(value, fmt) : date(value, fmt);
                }
                if (t != NT_DATE && tz)
                    value = date(value);
                return tz ? tz.date(value) : date(value);
            }
            default: {
              throw "FIELD-TYPE-ERROR", sprintf("output type %y not supported", type.type);
            }
        }
    }

    #! Validate and prepare global fixed-length options
    /**
        @param opts Options passed during class instantiation.
    */
    private *hash checkOptions(*hash opts) {
        string errname = 'RECORD-ITERATOR-ERROR';

        foreach my hash i in (opts.pairIterator()) {
            switch (i.key) {
                case "eol": {
                    if (i.value.typeCode() != NT_STRING) {
                        throw errname, sprintf("expecting a string value to option %y; got %y (type %s) instead", i.key, i.value, i.value.type());
                    }
                    if (!inlist(i.value, EOLS)) {
                        throw errname, sprintf("expecting a known end of line value to option %y; got %y (type %s) instead", i.key, i.value, i.value.type());
                    }
                    opts.eol = i.value;
                    break;
                }
                case "timezone": {
                    opts.timezone = new TimeZone(opts.timezone);
                    break;
                }
                case "ignore_empty": {
                    opts.ignore_empty = boolean(opts.ignore_empty);
                    break;
                }
                case "encoding": {
                    # TODO: check encoding name here
                    break;
                }
                case "number_format": {
                    # TODO: check here
                    break;
                }
                case "date_format": {
                    # TODO: check here
                    break;
                }
            }
        }

        return opts;
    }

    #! Identify a fixed-length line type using a second method %identifyTypeImpl, that may be overridden if necessary.
    /**
        @param input_line Input line to be identified

        @return the name of the record corresponding to the input line

        @throw FIXED-LENGTH-UTIL-NON-MATCHING-TYPE input line cannot be matched to a known record
    */
    string identifyType(string input_line) {
        *string type = identifyTypeImpl(input_line);
        if (!type || !m_specs.hasKey(type))
            throw "FIXED-LENGTH-UTIL-NON-MATCHING-TYPE", sprintf("Line identified to be of type that is not present in the spec '%s'", input_line);
        return type;
    }

    #! Identify a fixed-length line type, given the raw line string. This method performs a lookup to a precalculated table based on line lengths (see constructor()). In case a different criteria is needed, eg. when two line types in a spec have the same length, this method needs to be overridden, and will throw an exception, because the precalculated mapping will be empty.
    /**
        @param input_line Input line to be identified

        @return the record name or @ref nothing if the input cannot be matched

        @throw FIXED-LENGTH-UTIL-NON-MATCHING-TYPE input line cannot be matched to a known record or the input line's length does not match the expected length
    */
    *string identifyTypeImpl(string input_line) {
        if (m_conflicts{input_line.size()})
            throw "FIXED-LENGTH-UTIL-NON-MATCHING-TYPE", sprintf("Line with byte length %d was not automatically matched since the following records have this length: %y; you need to provide your own identifyTypeImpl() method (record: %s)", input_line.size(), m_conflicts{input_line.size()}, input_line);

        *string rv = m_line_lengths{input_line.size()};
        if (!rv)
            throw "FIXED-LENGTH-UTIL-NON-MATCHING-TYPE", sprintf("Line of unexpected length %d found; known lengths: %y (record: %s)", input_line.size(), (map $1.toInt(), m_line_lengths.keyIterator()), input_line);

        return rv;
    }

    #! A verification function to be called for each line. This method can be overridden to achieve a begin-to-end validation of the whole input file.
    /**
        @param from Type of previous line being processed
        @param to Type of the current line being processed
    */
    bool checkTransition(*string from, *string to) {
        return True;
    }
}

#! Structured line iterator for fixed-length line files allowing efficient "pipelined" processing.
/**
    @par Example:
    @code
#!/usr/bin/env qr

%requires FixedLengthUtil

hash specs = (
    "type1" : (
        "col1": ("length": 5, "type": "int"),
        "col2": ("length": 2, "type": "string"),
    ),
    "type2" : (
        "col3": ("length": 1, "type": "string"),
        "col4": ("length": 3, "type": "string"),
        "col5": ("length": 8, "type": "date", "format": "DDMMYYYY",
        #"timezone": "Europe/Prague", # use global if omitted
        ),
    ),
);

hash global_options = (
    "encoding"     : "UTF-8",
    "eol"          : "\n",
    "ignore_empty" : True,
    "timezone"     : "Europe/Prague", # used if not overridden in a date field specification
);

FixedLengthFileIterator i(file, specs, global_options);
# output all records
map printf("%y\n", $1), i;
    @endcode
*/
public class FixedLengthUtil::FixedLengthFileIterator inherits FixedLengthUtil::FixedLengthAbstractIterator, Qore::FileLineIterator {
    #! Instantiates the FixedLengthFileIterator object.
    /**
        @param file_name File path to read
        @param spec Fixed-length line specification; see @ref fixedlengthspec for more information
        @param opts Global options; see @ref fixedlengthglobals for more information
    */
    constructor(string file_name, hash spec, *hash opts) : FixedLengthUtil::FixedLengthAbstractIterator(spec, opts), FileLineIterator(file_name, opts.encoding.typeCode() == NT_STRING ? opts.encoding : NOTHING, opts.eol.typeCode() == NT_STRING ? opts.eol : NOTHING) {
        # do not convert every string to the same encoding
        m_opts.input_encoding = remove m_opts.encoding;
    }

    #! returns a line
    private string getLine() {
        return FileLineIterator::getValue();
    }

    #! Moves the current line / record position to the next line / record; returns @ref False if there are no more lines to iterate
    /** This method will return @ref True again after it returns @ref False once if the file being iterated has data that can be iterated, otherwise it will always return @ref False. The iterator object should not be used to retrieve a value after this method returns @ref False.
        @return @ref False if there are no lines / records to iterate (in which case the iterator object is invalid and should not be used); @ref True if successful (meaning that the iterator object is valid)

        @note that empty lines are ignored if "ignore_empty" option is in effect
     */
    bool next() {
        bool status = FileLineIterator::next();
        if (m_opts.ignore_empty) {
            while (status && getLine() == '') {
                status = FileLineIterator::next();
            }
        }
        return status;
    }
}

#! Structured line iterator for fixed-length line strings allowing efficient "pipelined" processing.
/**
    @par Example:
    @code
hash specs = (
    "type1" : (
        "col1": ("length": 5, "type": "int"),
        "col2": ("length": 2, "type": "string"),
    ),
    "type2": (
        "col3": ("length": 1, "type": "string"),
        "col4": ("length": 3, "type": "string"),
        "col5": ("length": 8, "type": "date", "format": "DDMMYYYY",
        # "timezone" : "Europe/Prague", # use global if omitted
        ),
    ),
);

hash global_options = (
    "encoding"     : "UTF-8",
    "eol"          : "\n",
    "ignore_empty" : True,
    "timezone"     : "Europe/Prague", # used if not overridden in a date field specification
);

FixedLengthDataIterator i(string, specs, global_options);
while (i.next()) {
    operation_with_hash(i.getValue())
}
    @endcode
*/
public class FixedLengthUtil::FixedLengthDataIterator inherits FixedLengthUtil::FixedLengthAbstractIterator, Qore::DataLineIterator {
    private {
        int position;
        int max_index;

        list lines;
        int lines_pos;

        string data;
    }

    #! Instantiates the FixedLengthFileIterator object
    /** @par Example:
        @code
hash specs = (
    "type1" : (
        "col1": ("length": 5, "type": "int"),
        "col2": ("length": 2, "type": "string"),
    ),
    "type2": (
        "col3": ("length": 1, "type": "string"),
        "col4": ("length": 3, "type": "string"),
        "col5": ("length": 8, "type": "date", "format": "DDMMYYYY",
        # "timezone" : "Europe/Prague", # use global if omitted
        ),
    ),
);

hash global_options = (
    "encoding"     : "UTF-8",
    "eol"          : "\n",
    "ignore_empty" : True,
    "timezone"     : "Europe/Prague", # used if not overridden in a date field specification
);

FixedLengthDataIterator i(string, specs, global_options);
while (i.next()) {
    operation_with_hash(i.getValue())
}
        @endcode

        @param input_data The input string to process
        @param spec Fixed-length line specification; see @ref fixedlengthspec for more information
        @param opts Global options; see @ref fixedlengthglobals for valid values
    */
    constructor(string input_data, hash spec, *hash opts) : FixedLengthUtil::FixedLengthAbstractIterator(spec, opts), DataLineIterator(input_data, opts.eol.typeCode() == NT_STRING ? opts.eol : NOTHING, opts.trim.typeCode() == NT_BOOLEAN ? opts.trim : NOTHING) {
        data  = input_data;
    }

    #! Moves the current line / record position to the next line / record; returns @ref False if there are no more lines to iterate
    /** This method will return @ref True again after it returns @ref False once if the file being iterated has data that can be iterated, otherwise it will always return @ref False. The iterator object should not be used to retrieve a value after this method returns @ref False.
        @return @ref False if there are no lines / records to iterate (in which case the iterator object is invalid and should not be used); @ref True if successful (meaning that the iterator object is valid)

        @note that empty lines are ignored if "ignore_empty" option is in effect
     */
    bool next() {
        any status = DataLineIterator::next();
        if (m_opts.ignore_empty) {
            while (status && DataLineIterator::getValue() == "") { # skip empty lines
                status = DataLineIterator::next();
            }
        }
        return status;
    }
}

#! Structured abstract writer for fixed-length lines
public class FixedLengthUtil::FixedLengthAbstractWriter {
    private {
        hash m_specs;
        string m_state;
        int m_linescount = 0;
        hash m_opts = {};
    }

    #! creates the FixedLengthAbstractWriter object
    /**
        @param specs Fixed-length line specification; see @ref fixedlengthspec for more information
        @param opts Global options; see @ref fixedlengthglobals for valid values
    */
    constructor(hash specs, *hash opts) {
        m_specs = specs;
        delete opts.eol;
        m_opts = m_opts + checkOptions(opts);

        # validate padding options
        HashIterator hit(specs);
        while (hit.next()) {
            HashIterator cit(hit.getValue());
            while (cit.next()) {
                if (cit.getValue(){"padding"}) {
                    if (cit.getValue(){"padding"} != "left" && cit.getValue(){"padding"} != "right")
                        throw "FIXED-LENGTH-UTIL-PADDING-ERROR", sprintf("record type: %s; field: %s; Key: padding; value must be 'left' or 'right'. Got: %s",
                                                                         hit.getKey(), cit.getKey(), cit.getValue(){"padding"});
                }
                if (cit.getValue(){"padding_char"} && cit.getValue(){"padding_char"}.length() != 1) {
                    throw "FIXED-LENGTH-UTIL-PADDING-ERROR", sprintf("record type: %s; field: %s; Key: padding; value must have a character length of 1; got: %d instead",
                                                                     hit.getKey(), cit.getKey(), cit.getValue(){"padding_char"}.length());
                }
            }
        }
    }

    private *hash checkOptions(*hash opts) {
        string errname = 'RECORD-WRITER-ERROR';

        foreach my hash i in (opts.pairIterator()) {
            switch (i.key) {
                case "eol": {
                    if (i.value.typeCode() != NT_STRING) {
                        throw errname, sprintf("expecting a string value to option %y; got %y (type %s) instead", i.key, i.value, i.value.type());
                    }
                    if (!inlist(i.value, EOLS)) {
                        throw errname, sprintf("expecting a known end of line value to option %y; got %y (type %s) instead", i.key, i.value, i.value.type());
                    }
                    opts.eol = i.value;
                    break;
                }
            }
        }

        return opts;
    }

    #! get processed lines count
    int linesCount() {
        return m_linescount;
    }

    #! Formats a single line from a hash describing the record type and its contents
    /**
        @param line a hash describing the record to be formatted; the hash must have the following keys:
        - \c "type": a string giving the record type (must be defined in @ref fixedlengthspec given in the constructor)
        - \c "record": a hash giving the input record to be rendered (with keys as defined in the @ref fixedlengthspecrecordhash for the record identified by the \a type argument)

        @throw INVALID-LINE-DATA line argument missing either \a type or \a record keys
        @throw INVALID-RECORD record name (\a type key in the record hash) not recognized
        @throw FIELD-INPUT-ERROR the input value is too large to render into the output field
        @throw RECORD-TRANSITION-ERROR a record transition error occurred; an invalid record sequence was given in the input data
    */
    string formatLine(hash line) {
        if (!line.type)
            throw "INVALID-LINE-DATA", sprintf("invalid line data; missing \"type\" key; got keys: %y", line.keys());
        if (!line.record)
            throw "INVALID-LINE-DATA", sprintf("invalid line data; missing \"record\" key; got keys: %y", line.keys());
        string type = line.type;
        if (!m_specs{type})
            throw "INVALID-RECORD", sprintf("unknown record type %y; valid records: %y", type, m_specs.keys());

        string result = "";
        foreach string field in (m_specs{type}.keyIterator()) {
            # field description
            hash fh = m_specs{type}{field};

            any value = line.record{field};
            # format output
            switch (fh.type) {
                case "float":
                case "number": {
                    *string fmt = fh.format ?? m_opts.number_format;
                    if (fmt)
                        value = format_number(fmt, value);
                    else if (value.typeCode() != NT_STRING)
                        value = string(value);
                    break;
                }
                case "date": {
                    switch (value.typeCode()) {
                        case NT_DATE: {
                            *string fmt = fh.format ?? m_opts.date_format;
                            value = fmt ? value.format(fmt) : string(value);
                            break;
                        }
                        case NT_STRING:
                            break;
                        default:
                            value = string(value);
                        break;
                    }
                }
                default:
                    if (value.typeCode() != NT_STRING)
                        value = string(value);
                    break;
            }

            int len = value.length();
            # Validate
            if (len > fh.length)
                throw "FIELD-INPUT-ERROR", sprintf("Value %y (len %d) too large to pack into field: %s (len %d)", value, len, field, fh.length);
            # do output justification
            int spacingsize = fh.length - len;
            if (spacingsize) {
                string spacing = strmul(fh.padding_char ?? " ", spacingsize);

                if (fh.padding == "left" || (!fh.padding && fh.type == "int"))
                    result += spacing + value;
                else
                    result += value + spacing;
            }
            else
                result += value;
        }

        if (!checkTransition(m_state, type)) {
            throw "RECORD-TRANSITION-ERROR", sprintf("record %y cannot follow record %y for line %y", (type ?? "<START>"), (m_state ?? "<START>"), result);
        }
        m_state = type;
        ++m_linescount;

        return result;
    }

    #! A verification function to be called for each line. This method can be overridden to achieve a begin-to-end validation of the whole input file.
    /**
        @param from Type of previous line being processed
        @param to Type of the current line being processed
    */
    bool checkTransition(*string from, *string to) {
        return True;
    }
}

#! Writer for fixed-length-line output files
/** @par Example:
    @code
#!/usr/bin/env qr

%requires FixedLengthUtil

list data = (
    ("type": "type1", "record": {"col1" : 11111, "col2" : "bb"}),
    ("type": "type2", "record": {"col3" : "c", "col4" : "ddd", "col5" : "31122014"}),
    ("type": "type1", "record": {"col1" : 22222, "col2" : "gg"}),
);

hash specs = (
    "type1" : (
        "col1": ("length": 5, "type": "int"),
        "col2": ("length": 2, "type": "string"),
    ),
    "type2" : (
        "col3" : ("length": 1, "type": "string"),
        "col4" : ("length" : 3, "type": "string"),
        "col5" : ("length" : 8, "type": "date", "format": "DDMMYYYY", "timezone" : "Europe/Prague"),
    ),
);

hash global_options = (
    "eol" : "\n",
);

FixedLengthFileWriter w(file, specs, global_options);
w.write(data);
    @endcode

    @see FixedLengthDataWriter
*/
public class FixedLengthUtil::FixedLengthFileWriter inherits FixedLengthUtil::FixedLengthAbstractWriter {
    private {
        string file;
        File m_file();
    }

    #! Instantiates the FixedLengthFileWriter object
    /**
        @param file_name the output file path
        @param specs Fixed-length line specification; see @ref fixedlengthspec for more information
        @param opts Global options; see @ref fixedlengthglobals for valid values
    */
    constructor(string file_name, hash specs, *hash opts) : FixedLengthUtil::FixedLengthAbstractWriter(specs, opts)  {
        file = file_name;
        m_opts.eol = m_opts.eol ?? "\n";
        m_file.open2(file, O_WRONLY | O_CREAT | m_opts.file_flags);
    }

    #! Return the file name
    string getFileName() {
        return file;
    }

    #! Renders a single line for a single input record hash to the output file
    /** @par Example:
        @code
const Spec = (
    "header": (
        "flow_type": ("length": 3, "type": "string"),
        "record_type": ("length": 3, "type": "int", "padding_char": "0"),
        "number_of_records": ("length": 8, "type": "int", "padding_char": "0"),
    ),
);
hash hh = (
    "type": "header",
    "record": (
        "flow_type": "WTS",
        "record_type": "950",
        "number_of_records": 1,
     ),
);
FixedLengthFileWriter w(Spec);
w.write(hh);
        @endcode

        @param line A hash representing input data to be writen to the file; the hash must have the following keys:
        - \c "type": a string giving the record type (must be defined in @ref fixedlengthspec given in the constructor)
        - \c "record": a hash giving the input record to be rendered (with keys as defined in the @ref fixedlengthspecrecordhash for the record identified by the \a type argument)

        @throw INVALID-RECORD record name (\a type key in the record hash) not recognized
        @throw FIELD-INPUT-ERROR the input value is too large to render into the output field
        @throw RECORD-TRANSITION-ERROR a record transition error occurred; an invalid record sequence was given in the input data
    */
    write(hash line) {
        m_file.write(formatLine(line) + m_opts.eol);
    }

    #! iterates the input records and writes rendered versions of all input records to the file
    /** @par Example:
        @code
FixedLengthFileWriter w(output_file, Spec);
w.write(i);
        @endcode

        @param lines An iterator to stream input records; each iterator value must be a hash with the following keys:
        - \c "type": a string giving the record type (must be defined in @ref fixedlengthspec given in the constructor)
        - \c "record": a hash giving the input record to be rendered (with keys as defined in the @ref fixedlengthspecrecordhash for the record identified by the \a type argument)

        @throw INVALID-RECORD record name (\a type key in the record hash) not recognized
        @throw FIELD-INPUT-ERROR the input value is too large to render into the output field
        @throw RECORD-TRANSITION-ERROR a record transition error occurred; an invalid record sequence was given in the input data
    */
    write(list lines) {
        map m_file.write(formatLine($1) + m_opts.eol), lines;
    }

    #! iterates the input records and writes rendered versions of all input records to the file
    /** @par Example:
        @code
FixedLengthFileWriter w(output_file, Spec);
w.write(i);
        @endcode

        @param lines An iterator to stream input records; each iterator value must be a hash with the following keys:
        - \c "type": a string giving the record type (must be defined in @ref fixedlengthspec given in the constructor)
        - \c "record": a hash giving the input record to be rendered (with keys as defined in the @ref fixedlengthspecrecordhash for the record identified by the \a type argument)

        @throw INVALID-RECORD record name (\a type key in the record hash) not recognized
        @throw FIELD-INPUT-ERROR the input value is too large to render into the output field
        @throw RECORD-TRANSITION-ERROR a record transition error occurred; an invalid record sequence was given in the input data
    */
    write(Qore::AbstractIterator lines) {
        map m_file.write(formatLine($1) + m_opts.eol), lines;
    }
}

#! Writer class for fixed-length-line data
/** @par Example:
    @code
const Spec = (
    "header": (
        "flow_type": ("length": 3, "type": "string"),
        "record_type": ("length": 3, "type": "int", "padding_char": "0"),
        "number_of_records": ("length": 8, "type": "int", "padding_char": "0"),
    ),
    "line": (
        "flow_type": ("length": 3, "type": "string"),
        "record_type": ("length": 3, "type": "int", "padding_char": "0"),
        "processing_id": ("length": 10, "type": "int", "padding_char": "0"),
        "processing_name": ("length": 10, "type": "string"),
        "po_number": ("length": 10, "type": "int", "padding_char": "0"),
    ),
    "trailer": (
        "flow_type": ("length": 3, "type": "string"),
        "record_type": ("length": 3, "type": "int", "padding_char": "0"),
        "number_of_records": ("length": 8, "type": "int", "padding_char": "0"),
    ),
);
FixedLengthDataWriter w(Spec);
AbstractIterator i = get_record_iterator();
string data = w.write(i);
    @endcode

    @see FixedLengthFileWriter
*/
public class FixedLengthUtil::FixedLengthDataWriter inherits FixedLengthUtil::FixedLengthAbstractWriter {
    #! Instantiates the FixedLengthDataWriter object.
    /**
        @param specs Fixed-length line specification; see @ref fixedlengthspec for more information
        @param opts Global formatting options; see @ref fixedlengthglobals for valid values
    */
    constructor(hash specs, *hash opts) : FixedLengthUtil::FixedLengthAbstractWriter(specs, opts) {
    }

    #! iterates the input records and returns a string corresponding to the rendered versions of all input records
    /** @par Example:
        @code
FixedLengthDataWriter w(Spec);
string file = w.write(input_list);
        @endcode

        @param lines An iterator to stream input records; each iterator value must be a hash with the following keys:
        - \c "type": a string giving the record type (must be defined in @ref fixedlengthspec given in the constructor)
        - \c "record": a hash giving the input record to be rendered (with keys as defined in the @ref fixedlengthspecrecordhash for the record identified by the \a type argument)

        @return the rendered string for all input records

        @throw INVALID-RECORD record name (\a type key in the record hash) not recognized
        @throw FIELD-INPUT-ERROR the input value is too large to render into the output field
        @throw RECORD-TRANSITION-ERROR a record transition error occurred; an invalid record sequence was given in the input data
    */
    string write(list lines) {
        string result = "";
        string eol = m_opts.eol ? m_opts.eol : "\n";
        map result += formatLine($1) + eol, lines;
        return result;
    }

    #! Return a single line for a single input record hash
    /** @par Example:
        @code
const Spec = (
    "header": (
        "flow_type": ("length": 3, "type": "string"),
        "record_type": ("length": 3, "type": "int", "padding_char": "0"),
        "number_of_records": ("length": 8, "type": "int", "padding_char": "0"),
    ),
);
hash hh = (
    "type": "header",
    "record": (
        "flow_type": "WTS",
        "record_type": "950",
        "number_of_records": 1,
     ),
);
FixedLengthDataWriter w(Spec);
string line = w.write(hh);
        @endcode

        @param line A hash representing input data to be rendered as a string; the hash must have the following keys:
        - \c "type": a string giving the record type (must be defined in @ref fixedlengthspec given in the constructor)
        - \c "record": a hash giving the input record to be rendered (with keys as defined in the @ref fixedlengthspecrecordhash for the record identified by the \a type argument)

        @return a single line for a single input record hash

        @throw INVALID-RECORD record name (\a type key in the record hash) not recognized
        @throw FIELD-INPUT-ERROR the input value is too large to render into the output field
        @throw RECORD-TRANSITION-ERROR a record transition error occurred; an invalid record sequence was given in the input data
    */
    string write(hash line) {
        return formatLine(line) + (m_opts.eol ? m_opts.eol : "\n");
    }

    #! iterates the input records and returns a string corresponding to the rendered versions of all input records
    /** @par Example:
        @code
FixedLengthDataWriter w(Spec);
string file = w.write(i);
        @endcode

        @param lines An iterator to stream input records; each iterator value must be a hash with the following keys:
        - \c "type": a string giving the record type (must be defined in @ref fixedlengthspec given in the constructor)
        - \c "record": a hash giving the input record to be rendered (with keys as defined in the @ref fixedlengthspecrecordhash for the record identified by the \a type argument)

        @return the rendered string for all input records

        @throw INVALID-RECORD record name (\a type key in the record hash) not recognized
        @throw FIELD-INPUT-ERROR the input value is too large to render into the output field
        @throw RECORD-TRANSITION-ERROR a record transition error occurred; an invalid record sequence was given in the input data
    */
    string write(Qore::AbstractIterator lines) {
        string result = "";
        string eol = m_opts.eol ? m_opts.eol : "\n";
        map result += formatLine($1) + eol, lines;
        return result;
    }
}