System Initiative

// Converts a PCRE pattern from the cloudformation schema into a JS RegExp // This is not intended to exhaustively convert any PCRE to RegExp; it is designed to be // simple and Good Enough (TM) to handle the published AWS CF schemas. export function cfPcreToRegexp(pattern: string) { if (pattern in BROKEN_REGEXES) return BROKEN_REGEXES[pattern]; // Extract flags from the beginning of the pattern let flags = ""; const matchFlags = pattern.match(/^\(\?([ims]+)\)/); if (matchFlags) { flags += matchFlags[1]; pattern = pattern.slice(matchFlags[0].length); } // Go through, looking for ?-style subpatterns we can replace const replacements = []; for (let index = 0; index < pattern.length; index++) { switch (pattern[index]) { case "(": // We only care about special patterns. All else is fine. if (pattern[index + 1] === "?") { switch (pattern[index + 2]) { // (?:, ?=, ?!, ?<= and ?<! translate just fine) case ":": case "=": case "!": case "<": break; case "-": if (pattern.slice(index).startsWith("(?-s:.*)")) { replacements.push({ start: index, end: index + "(?-s:.*)".length, replacement: "[^\\r\\n]*", }); index += "(?-s:.*)".length - 1; } else { throw new Error( `Unsupported subpattern: ${pattern.slice(index)}`, ); } break; default: throw new Error( `Unsupported subpattern: ${pattern.slice(index)}`, ); } } break; // Skip an extra character if it's an escape code case "\\": switch (pattern[index + 1]) { // \A and \Z match the start and end of a string, respectively case "A": replacements.push({ start: index, end: index + 2, replacement: "^", }); break; case "Z": replacements.push({ start: index, end: index + 2, replacement: "$", }); break; } index++; break; case "[": // Skip stuff in [] (could be a parenthesis) while (index < pattern.length && pattern[index] !== "]") { if (pattern[index] === "\\") { switch (pattern[index + 1]) { // If \p is specified, the "u" flag is required. This makes things more strict, so // we don't enable it for everything ... case "p": if (!flags.includes("u")) flags += "u"; break; } index++; } index++; } break; } } // Go back through in reverse order, replacing anything that needs it while (replacements.length > 0) { const { start, end, replacement } = replacements.pop()!; pattern = pattern.substring(0, start) + replacement + pattern.substring(end); } return { pattern, flags: flags === "" ? undefined : flags }; } const BROKEN_REGEXES: Record< string, { pattern: string; flags?: string } | undefined > = { // The * is redundant, here--it's fairly clear the intent was to limit to 1-1000 characters "^[a-zA-Z-0-9-:\\/]*{1,1000}$": { pattern: "^[a-zA-Z-0-9-:\\/]{1,1000}$", flags: undefined, }, "^[a-zA-Z-0-9-:\\/.]*{1,1000}$": { pattern: "^[a-zA-Z-0-9-:\\/.]{1,1000}$", flags: undefined, }, "[\\u0020-\\uD7FF\\uE000-\\uFFFD\\uD800\\uDC00-\\uDBFF\\uDFFF\t]*": { pattern: "[\\u0020-\\uD7FF\\uE000-\\uFFFD\\uD800\\uDFFF\t]*", flags: undefined, }, "[\\u0020-\\uD7FF\\uE000-\\uFFFD\\uD800\\uDC00-\\uDBFF\\uDFFF\\t]*": { pattern: "[\\u0020-\\uD7FF\\uE000-\\uFFFD\\uD800\\uDFFF\\t]*", flags: undefined, }, "^(?! )[\\p{L}\\p{N}\\p{Z}-_]*(?<! )$": { pattern: "^(?! )[\\p{L}\\p{N}\\p{Z}\\-_]*(?<! )$", flags: undefined, }, "^(?!(?i)aws)[A-Za-z0-9]{2,64}::[A-Za-z0-9]{2,64}::[A-Za-z0-9]{2,64}$": { pattern: "^(?!aws)[A-Za-z0-9]{2,64}::[A-Za-z0-9]{2,64}::[A-Za-z0-9]{2,64}$", flags: "i", }, // I don't even "[ -퟿-�𐀀-􏿿\r\n\t]*": undefined, // This doesn't compile--says it's an invalid property. Seems unsupported. "[\\p{Graph}\\x20]*": undefined, };