First iteration on number formatting

pkg/i18n/number_format.dart

Index: pkg/i18n/number_format.dart
===================================================================
--- pkg/i18n/number_format.dart	(revision 0)
+++ pkg/i18n/number_format.dart	(revision 0)
@@ -0,0 +1,258 @@
+// Copyright (c) 2012, the Dart project authors.  Please see the AUTHORS file
+// for details. All rights reserved. Use of this source code is governed by a
+// BSD-style license that can be found in the LICENSE file.
+
+#library("number_format");

[Emily Fortuna, 2012/08/15 21:38:16]

nit: space between library and imports section?

[Alan Knight, 2012/08/15 23:57:44]

Done.

+#import("intl.dart");
+#import("number_symbols.dart");
+#import("number_symbols_data.dart");
+
+class NumberFormat {
+

[Emily Fortuna, 2012/08/15 21:38:16]

remove extra newline

[Alan Knight, 2012/08/15 23:57:44]

Done.

+  /**
+   * Create a number format that prints in [newPattern] as it applies in
+   * [locale].
+   */
+  NumberFormat([String newPattern, String locale]) {

[Emily Fortuna, 2012/08/15 21:38:16]

Can you add a TODO around here reminding us this API needs to change to be
async?

[Alan Knight, 2012/08/15 23:57:44]

Done.

+    _locale = Intl.verifiedLocale(locale);
+    _setPattern(newPattern);
+  }
+
+  /** Variables to determine how number printing behaves. */
+  String _negativePrefix = '-';

[Emily Fortuna, 2012/08/15 21:38:16]

put fields above the constructor.

[Alan Knight, 2012/08/15 23:57:44]

Done.

+  String _positivePrefix = '';
+  String _negativeSuffix = '';
+  String _positiveSuffix = '';
+  num _multiplier = 1;
+  num _groupingSize = 3;

[Emily Fortuna, 2012/08/15 21:38:16]

what's this one mean? add a comment?

[Alan Knight, 2012/08/15 23:57:44]

Done.

+  bool _decimalSeparatorAlwaysShown = false;
+  bool _useExponentialNotation = false;
+  int _maximumIntegerDigits = 40;

[Emily Fortuna, 2012/08/15 21:38:16]

how are these determined?

[Alan Knight, 2012/08/15 23:57:44]

These are hard-coded versions appropriate for an en_US type locale. Ultimately I
expect them to come from ICU data for a particular locale. They may also
disappear as variables and turn into references into the locale data, but that's
not clear.

+  int _minimumIntegerDigits = 1;

[Emily Fortuna, 2012/08/15 21:38:16]

should these be final? 

[Alan Knight, 2012/08/15 23:57:44]

They will depend on the locale. At least with the current implementation I don't
think they can be final if they rely on the result of calling a method
(verifiedLocale()).

[Emily Fortuna, 2012/08/16 00:10:24]

What about declaring the variable as final and initializing in the constructor
after verifiedLocale is called? (or is that not a legal use of final? I can
never remember)

[Alan Knight, 2012/08/16 16:41:49]

Hmm. I didn't think it was, but now it looks like it is, but could be tricky. To
be final, I need to initialize them in the initializer list, not in the
constructor body, but they will depend on both the locale and the pattern, so
we'd have to call the logic repeatedly to parse those, or else we'd need to make
this a factory constructor. That seems complicated, and these might disappear as
variables anyway, so just made it a TODO to see if they can be made final.

+  int _maximumFractionDigits = 3; // invariant, >= minFractionDigits
+  int _minimumFractionDigits = 0;
+  int _minimumExponentDigits = 0;
+  bool _useSignForPositiveExponent = false;
+
+  /** The locale in which we print numbers. */
+  String _locale;
+
+  /** Caches the symbols used for our locale. */
+  NumberSymbols _symbols;
+
+  /**
+   * Transient internal state in which to build up the result of the format
+   *  operation. We can have this be just an instance variable because Dart is

[Emily Fortuna, 2012/08/15 21:38:16]

indentation

[Alan Knight, 2012/08/15 23:57:44]

Done.

+   *  single-threaded and unless we do an asynchronous operation in the process
+   *  of formatting then there will only ever be one number being formatted
+   *  at a time. In languages with threads we'd need to pass this on the stack.

[Emily Fortuna, 2012/08/15 21:38:16]

Is this comment relevant? We don't have threads.....

[Alan Knight, 2012/08/15 23:57:44]

I'm saying we can do this precisely because we don't have threads, so it's a
pattern that works in Dart but would be very bad practice in Java, C++, etc. I
thought it needed explaining so as to show the contrast.

[Emily Fortuna, 2012/08/16 00:10:24]

Gotcha. I wasn't reading this carefully.

+   */
+  StringBuffer _buffer;
+
+  /**
+   * Return the locale code in which we operate, e.g. 'en_US' or 'pt'.
+   */
+  String get locale() => _locale;
+
+  /**
+   * Return the symbols which are used in our locale. Cache them to avoid
+   * repeated lookup.
+   */
+  NumberSymbols get symbols() {
+    if (_symbols == null) _symbols = numberFormatSymbols[locale];
+    return _symbols;
+  }
+
+  // TODO(alanknight): Actually use the pattern and locale.
+  _setPattern(x) {}

[Emily Fortuna, 2012/08/15 21:38:16]

type for arg x?

[Alan Knight, 2012/08/15 23:57:44]

Done.

+
+  /**
+   * Format [number] according to our pattern and return the formatted string.
+   */
+  String format(num number) {
+    if (number.isNaN()) return symbols.NAN;
+    if (number.isInfinite()) return "${_signPrefix(number)}${symbols.INFINITY}";
+
+    _newBuffer();
+    _add(_signPrefix(number));
+    _formatNumber(number.abs());
+    _add(_signSuffix(number));
+
+    var result = _buffer.toString();
+    _buffer = null;
+    return result;
+  }
+
+  /**
+   * Format the main part of the number in the form dictated by the pattern.
+   */
+  void _formatNumber(num number) {
+    if (_useExponentialNotation) {
+        _formatExponential(number);

[Emily Fortuna, 2012/08/15 21:38:16]

indentation :-(

[Alan Knight, 2012/08/15 23:57:44]

Done.

+    } else {
+        _formatFixed(number);
+    }
+  }
+
+  /** Format the number in exponential notation. */
+  _formatExponential(num number) {
+    if (number == 0.0) {
+      _formatFixed(number);
+      _formatExponent(0);
+      return;
+    }
+
+    var exponent = (Math.log(number) / Math.log(10)).floor();
+    var mantissa = number / Math.pow(10, exponent);
+
+    if (_minimumIntegerDigits < 1) {
+      exponent++;
+      mantissa /= 10;
+    } else {
+      exponent -= _minimumIntegerDigits - 1;
+      mantissa *= Math.pow(10, _minimumIntegerDigits - 1);
+    }
+    _formatFixed(number);
+    _formatExponent(exponent);
+  }
+
+  /**
+   * Format the exponent portion, e.g. in "1.3e-5" the "e-5".
+   */
+  void _formatExponent(num exponent) {
+    _add(symbols.EXP_SYMBOL);
+    if (exponent < 0) {
+      exponent = -exponent;
+      _add(symbols.MINUS_SIGN);
+    } else if (_useSignForPositiveExponent) {
+      _add(symbols.PLUS_SIGN);
+    }
+    _pad(_minimumExponentDigits, exponent.toString());
+  }
+
+  /**
+   * Format the basic number portion, inluding the digits after the decimal

[Emily Fortuna, 2012/08/15 21:38:16]

"digits after the decimal point" --> "fractional digits" (europe uses the ,)

[Alan Knight, 2012/08/15 23:57:44]

Done.

+   * point.
+   */
+  void _formatFixed(num number) {
+    // round the number

[Emily Fortuna, 2012/08/15 21:38:16]

Comments start with a capital letter and end in a period.

[Alan Knight, 2012/08/15 23:57:44]

Done.

+    var power = Math.pow(10, _maximumFractionDigits);
+    var intValue = number.truncate().toInt();
+    var multiplied = (number * power).round();
+    var fracValue = (multiplied - intValue * power).floor().toInt();
+

[Emily Fortuna, 2012/08/15 21:38:16]

is this an extra newline here?

[Alan Knight, 2012/08/15 23:57:44]

Done.

+    var fractionPresent = _minimumFractionDigits > 0 || fracValue > 0;
+
+    // On dartj2s the integer part may be large enough to be a floating

[Emily Fortuna, 2012/08/15 21:38:16]

but we just populated intValue = number.truncate().toInt(). So intValue *must*
be a dart int. could could it be larger...?

[Alan Knight, 2012/08/15 23:57:44]

Sadly, that's not true on dart2js. Like, seriously.
http://code.google.com/p/dart/issues/detail?id=1533

+    // point value, in which case we reduce it until it is small enough
+    // to be printed as an integer and pad the remainder with zeros.
+    var paddingDigits = new StringBuffer();
+    while (intValue is! int) {
+      paddingDigits.add(symbols.ZERO_DIGIT);
+      intValue = (intValue / 10).toInt();
+    }
+    var integerDigits = "${intValue}${paddingDigits}".charCodes();
+    var digitLength = integerDigits.length;
+
+    if (_hasPrintableIntegerPart(intValue)) {
+      _pad(_minimumIntegerDigits - digitLength);
+      for (var i = 0; i < digitLength; i++) {
+        _addDigit(integerDigits[i]);

[Emily Fortuna, 2012/08/15 21:38:16]

question -- are numbers always RTL, or are there some LTR? (aka should we  be
starting at the front of intengerDigits?)

[Alan Knight, 2012/08/15 23:57:44]

Good question. Added a TODO. At the moment this mirrors Closure code which
doesn't seem to do anything about printing them bidi, but that may be a
desirable thing.

+        _group(digitLength, i);
+      }
+    } else if (!fractionPresent) {
+      // If neither fraction nor integer part exists, just print zero.
+      _addZero();
+    }
+
+    _decimalSeparator(fractionPresent);
+    _formatFractionPart((fracValue + power).toString());
+  }
+
+  /**
+   * Format the part after the decimal place in a fixed point number.
+   */
+  void _formatFractionPart(String fractionPart) {
+    var fractionCodes = fractionPart.charCodes();
+    var fractionLength = fractionPart.length;
+    while (fractionPart[fractionLength - 1] == '0' &&
+           fractionLength > _minimumFractionDigits + 1) {
+      fractionLength--;
+    }
+    for (var i = 1; i < fractionLength; i++) {
+      _addDigit(fractionCodes[i]);
+    }
+  }
+
+  /** Print the decimal separator if appropriate. */
+  void _decimalSeparator(bool fractionPresent) {
+    if (_decimalSeparatorAlwaysShown || fractionPresent) {
+      _add(symbols.DECIMAL_SEP);
+    }
+  }
+
+  /**
+   * Return true if we have a main integer part which is printable, either
+   * because we have digits left of the decimal point, or because there are
+   * a minimum number of printable digits greater than 1.
+   */
+  bool _hasPrintableIntegerPart(int intValue) => intValue > 0 ||

[Emily Fortuna, 2012/08/15 21:38:16]

if the function is more than one line, make it use {}

[Alan Knight, 2012/08/15 23:57:44]

Done.

+      _minimumIntegerDigits > 0;
+
+  /**
+   * Create a new empty buffer. See comment on [_buffer] variable for why
+   * we have it as an instance variable rather than passing it on the stack.
+   */
+  void _newBuffer() { _buffer = new StringBuffer();}

[Emily Fortuna, 2012/08/15 21:38:16]

space after ;

[Alan Knight, 2012/08/15 23:57:44]

Done.

+
+  /** A group of methods that provide support for writing digits and other
+   * required characters into [_buffer] easily.
+   */
+  void _add(String x) { _buffer.add(x);}

[Emily Fortuna, 2012/08/15 21:38:16]

space after ; , here and below

[Alan Knight, 2012/08/15 23:57:44]

Done.

+  void _addCharCode(int x) { _buffer.addCharCode(x);}
+  void _addZero() { _buffer.add(symbols.ZERO_DIGIT);}
+  void _addDigit(int x) { _buffer.addCharCode(_localeZero + x - _zero);}
+
+  /** Print padding up to [numberOfDigits] above what's included in [basic]. */
+  void _pad(int numberOfDigits, [String basic = '']) {
+    for (var i = 0; i < numberOfDigits - basic.length; i++) {
+      _add(symbols.ZERO_DIGIT);
+    }
+    for (var x in basic.charCodes()) {
+      _addDigit(x);
+    }
+  }
+
+  /**
+   * Print the grouping symbol (e.g. comma as thousands separator) for digits
+   * left of the decimal place if it's required at this position.
+   */
+  String _group(int digitLen, int position) {

[Emily Fortuna, 2012/08/15 21:38:16]

what do digitLen vs position actually mean?

[Alan Knight, 2012/08/15 23:57:44]

Yes, that method was confusing. Rewrote and improved the comment.

+    if (digitLen - position > 1 && _groupingSize > 0 &&
+        ((digitLen - position) % _groupingSize == 1)) {
+      _add(symbols.GROUP_SEP);
+    }
+  }
+
+  /** Returns the code point for the character '0'. */
+  int get _zero() => '0'.charCodes()[0];
+
+  /** Returns the code point for the locale's zero digit. */
+  int get _localeZero() => symbols.ZERO_DIGIT.charCodeAt(0);
+
+  /**
+   * Returns the prefix for [x] based on wether it's positive or negative.

[Emily Fortuna, 2012/08/15 21:38:16]

wether -> whether

[Alan Knight, 2012/08/15 23:57:44]

Done.

+   * In en_US this would be '' and '-' respectively.
+   */
+  String _signPrefix(num x) {
+    return x.isNegative() ? _negativePrefix : _positivePrefix;
+  }
+
+  /**
+   * Returns the suffix for [x] based on wether it's positive or negative.
+   * In en_US there are no suffixes for positive or negative.
+   */
+  String _signSuffix(num x) {
+    return x.isNegative() ? _negativeSuffix : _positiveSuffix;
+  }
+}