Close the remaining documented parity edges: - Canonical XML 2.0 TrimTextNodes: an option that strips leading and trailing whitespace from each text node (dropping all-whitespace nodes), exposed as Options.canonical2. - XML Catalog delegation and chaining: delegatePublic, delegateSystem, delegateURI, and nextCatalog entries are parsed, and resolveSystem/ resolvePublic/resolveURI gain a loadingCatalog overload that follows the longest matching delegate prefix and then each nextCatalog, with a visited-set guard against cycles. - XPath number formatting: the canonical string form now expands any exponent in the shortest double rendering to a plain decimal, since XPath 1.0 forbids scientific notation. Closes #70
127 lines
5.8 KiB
Swift
127 lines
5.8 KiB
Swift
public extension PureXML.XPath {
|
|
/// An XPath 1.0 value: one of the four types the expression language operates
|
|
/// on. Every expression evaluates to one of these, and the conversion methods
|
|
/// implement the spec's coercion rules used by operators and functions.
|
|
/// A runtime value (a node-set holds live tree nodes), not a `Sendable`.
|
|
enum Value: Equatable {
|
|
case nodeSet([Node])
|
|
case boolean(Bool)
|
|
case number(Double)
|
|
case string(String)
|
|
|
|
/// The boolean() coercion: a node-set is true when non-empty, a number
|
|
/// when non-zero and not NaN, a string when non-empty.
|
|
public var boolean: Bool {
|
|
switch self {
|
|
case let .nodeSet(nodes): !nodes.isEmpty
|
|
case let .boolean(value): value
|
|
case let .number(value): value != 0 && !value.isNaN
|
|
case let .string(value): !value.isEmpty
|
|
}
|
|
}
|
|
|
|
/// The number() coercion: booleans map to 1/0, strings parse per the XPath
|
|
/// grammar (else NaN), and a node-set converts through its string-value.
|
|
public var number: Double {
|
|
switch self {
|
|
case .nodeSet: Self.parseNumber(string)
|
|
case let .boolean(value): value ? 1 : 0
|
|
case let .number(value): value
|
|
case let .string(value): Self.parseNumber(value)
|
|
}
|
|
}
|
|
|
|
/// The string() coercion: booleans render as `true`/`false`, numbers in the
|
|
/// XPath canonical form, and a node-set as the string-value of its first
|
|
/// node in document order (empty when the set is empty).
|
|
public var string: String {
|
|
switch self {
|
|
case let .nodeSet(nodes):
|
|
guard let first = nodes.min(by: Node.precedes) else { return "" }
|
|
return first.stringValue
|
|
case let .boolean(value): return value ? "true" : "false"
|
|
case let .number(value): return Self.format(value)
|
|
case let .string(value): return value
|
|
}
|
|
}
|
|
|
|
/// The node-set, or nil when the value is not a node-set.
|
|
public var nodes: [Node]? {
|
|
guard case let .nodeSet(nodes) = self else { return nil }
|
|
return nodes
|
|
}
|
|
|
|
/// Formats a number in the XPath canonical form: `NaN`, `Infinity`, an
|
|
/// integer without a decimal point, or the shortest round-tripping decimal.
|
|
/// XPath 1.0 forbids exponential notation, so any exponent in the shortest
|
|
/// representation is expanded to a plain decimal.
|
|
static func format(_ value: Double) -> String {
|
|
if value.isNaN { return "NaN" }
|
|
if value.isInfinite { return value < 0 ? "-Infinity" : "Infinity" }
|
|
if value == value.rounded(), abs(value) < 1e15 {
|
|
return String(Int64(value))
|
|
}
|
|
return expandExponent(String(value))
|
|
}
|
|
|
|
/// Rewrites a Swift double rendering into a plain decimal without an
|
|
/// exponent: `1.5e-05` becomes `0.000015`, `1.23e+21` becomes
|
|
/// `1230000000000000000000`.
|
|
private static func expandExponent(_ text: String) -> String {
|
|
let lowered = text.lowercased()
|
|
guard let exponentIndex = lowered.firstIndex(of: "e") else { return text }
|
|
let mantissa = String(lowered[lowered.startIndex ..< exponentIndex])
|
|
guard let exponent = Int(lowered[lowered.index(after: exponentIndex)...]) else { return text }
|
|
let negative = mantissa.hasPrefix("-")
|
|
let unsigned = negative ? String(mantissa.dropFirst()) : mantissa
|
|
let parts = unsigned.split(separator: ".", maxSplits: 1, omittingEmptySubsequences: false)
|
|
let integerDigits = String(parts.first ?? "")
|
|
let fractionDigits = parts.count > 1 ? String(parts[1]) : ""
|
|
var digits = Array(integerDigits + fractionDigits)
|
|
// The decimal point currently sits after `integerDigits`; the exponent
|
|
// shifts it right (positive) or left (negative).
|
|
var point = integerDigits.count + exponent
|
|
if point <= 0 {
|
|
digits = Array(repeating: "0", count: 1 - point) + digits
|
|
point = 1
|
|
} else if point > digits.count {
|
|
digits += Array(repeating: "0", count: point - digits.count)
|
|
}
|
|
let integerPart = String(digits[0 ..< point])
|
|
let fractionPart = String(digits[point...])
|
|
let body = fractionPart.isEmpty ? integerPart : "\(integerPart).\(trimTrailingZeros(fractionPart))"
|
|
return (negative ? "-" : "") + body
|
|
}
|
|
|
|
private static func trimTrailingZeros(_ fraction: String) -> String {
|
|
var trimmed = fraction
|
|
while trimmed.count > 1, trimmed.hasSuffix("0") {
|
|
trimmed.removeLast()
|
|
}
|
|
return trimmed
|
|
}
|
|
|
|
/// Parses a string to a number per the XPath `Number` grammar: optional
|
|
/// surrounding whitespace, an optional sign, and digits with an optional
|
|
/// single decimal point. Anything else is NaN.
|
|
static func parseNumber(_ raw: String) -> Double {
|
|
let trimmed = raw.trimmingXMLWhitespace()
|
|
guard !trimmed.isEmpty else { return .nan }
|
|
var seenDot = false
|
|
var digits = 0
|
|
for (offset, character) in trimmed.enumerated() {
|
|
if character == "-", offset == 0 { continue }
|
|
if character == "." {
|
|
if seenDot { return .nan }
|
|
seenDot = true
|
|
continue
|
|
}
|
|
guard character.isNumber else { return .nan }
|
|
digits += 1
|
|
}
|
|
guard digits > 0 else { return .nan }
|
|
return Double(trimmed) ?? .nan
|
|
}
|
|
}
|
|
}
|