hiperf_txt_parser 1.0.8 → 1.0.10

package/dist/index.d.ts

@@ -1,7 +1,7 @@
 export { parsePerfData, filterByTgid } from "./parser.js";
 export { formatPerfDataToText, formatPerfDataToJson } from "./serializer.js";
 export { toBackTraceStack, toBackTraceStacks } from "./backtrace.js";
-export { parseTraceFormat, parseCommonFieldsFromRaw, parseAllFieldsFromRaw, rawHexLinesToBuffer, buildTraceParserRegistry, decodeRawByRegistry, decodePerfRawData, } from "./traceFormat.js";
-export type { ParsedTraceFormat, TraceFormatField, TraceParserRegistry, DecodedRawSample, } from "./traceFormat.js";
+export { parseTraceFormat, parseCommonFieldsFromRaw, parseAllFieldsFromRaw, rawHexLinesToBuffer, bufferToRawHexLines, buildTraceParserRegistry, decodeRawByRegistry, decodePerfRawData, } from "./traceFormat.js";
+export type { ParsedTraceFormat, TraceFormatField, TraceParserRegistry, DecodedRawSample, DecodePerfRawDataOptions, Endian, } from "./traceFormat.js";
 export type { PerfData, RecordSample } from "./types.js";
 export type { RecordSampleJsonExportItem } from "./serializer.js";

package/dist/index.js

@@ -1,4 +1,4 @@
 export { parsePerfData, filterByTgid } from "./parser.js";
 export { formatPerfDataToText, formatPerfDataToJson } from "./serializer.js";
 export { toBackTraceStack, toBackTraceStacks } from "./backtrace.js";
-export { parseTraceFormat, parseCommonFieldsFromRaw, parseAllFieldsFromRaw, rawHexLinesToBuffer, buildTraceParserRegistry, decodeRawByRegistry, decodePerfRawData, } from "./traceFormat.js";
+export { parseTraceFormat, parseCommonFieldsFromRaw, parseAllFieldsFromRaw, rawHexLinesToBuffer, bufferToRawHexLines, buildTraceParserRegistry, decodeRawByRegistry, decodePerfRawData, } from "./traceFormat.js";

package/dist/traceFormat.d.ts

@@ -30,6 +30,11 @@ export interface DecodedRawSample {
     renderedText?: string;
     skipped: boolean;
 }
+export interface DecodePerfRawDataOptions {
+    /** 是否在替换后的 raw 内容中保留 common_* 信息 */
+    keepCommonFields?: boolean;
+}
+export type Endian = "le" | "be";
 /**
  * 解析 sample/trace_format 风格的文本，得到字段列表
  */
@@ -41,13 +46,23 @@ export declare function parseCommonFieldsFromRaw(raw: Uint8Array, format: Parsed
 /**
  * 解析 format 中全部字段（含数组），数组字段名会去掉 []，如 args[6] => args: [...]
  */
-export declare function parseAllFieldsFromRaw(raw: Uint8Array, format: ParsedTraceFormat): Record<string, number | bigint | Array<number | bigint>>;
+export declare function parseAllFieldsFromRaw(raw: Uint8Array, format: ParsedTraceFormat): Record<string, number | bigint | string | Array<number | bigint>>;
 /**
- * 将 perf 文本里 raw 段的 hex 行拼成连续字节（小端：每行一个数值，宽度由 hex 位数决定，4→2 字节，8→4 字节，16→8 字节）
+ * 将 perf 文本里 raw 段的 hex 行拼成连续字节（小端）。
+ *
+ * 规则：每行视为一个整数 token（如 0x12345678），按该 token 的字节宽度转成 LE 字节序。
+ * 例如：0x12345678 => [0x78, 0x56, 0x34, 0x12]
  */
 export declare function rawHexLinesToBuffer(lines: Array<{
     hex: string;
-}>): Uint8Array;
+}>, endian?: Endian): Uint8Array;
+/**
+ * 将字节缓冲按小端编码为 raw hex 行。
+ * 默认每行 4 字节（即 8 hex digits）。
+ */
+export declare function bufferToRawHexLines(raw: Uint8Array, bytesPerLine?: number, endian?: Endian): Array<{
+    hex: string;
+}>;
 /**
  * 从多个 trace_format 文本构建解析器集合（按 event ID 索引）
  */
@@ -67,4 +82,4 @@ export declare function decodeRawByRegistry(raw: Uint8Array, registry: TracePars
  * - 若找到 common_type 对应解析器：将 print fmt 渲染结果写入 `sample.raw.lines`（一行）。
  * - 若找不到解析器：放弃解析，并将 `common_type` 写入 `sample.raw.lines`（一行）。
  */
-export declare function decodePerfRawData(perfData: PerfData, registry: TraceParserRegistry): PerfData;
+export declare function decodePerfRawData(perfData: PerfData, registry: TraceParserRegistry, options?: DecodePerfRawDataOptions): PerfData;

package/dist/traceFormat.js

@@ -163,9 +163,38 @@ export function parseCommonFieldsFromRaw(raw, format) {
 export function parseAllFieldsFromRaw(raw, format) {
     const view = new DataView(raw.buffer, raw.byteOffset, raw.byteLength);
     const out = {};
+    const decoder = new TextDecoder("utf-8", { fatal: false });
     for (const field of format.fields) {
         const arr = parseArrayName(field.name);
         const key = arr ? arr.baseName : field.name;
+        const t = normalizeType(field.typeName);
+        // __data_loc char[] reason: 实际存一个 u32，低 16 位为 offset，高 16 位为 length
+        if (t.includes("__data_loc") && t.includes("char")) {
+            const locField = {
+                ...field,
+                typeName: "unsigned int",
+                signed: false,
+                size: 4,
+            };
+            const loc = readFieldScalar(view, locField);
+            if (typeof loc === "number") {
+                const locKey = `__data_loc_${field.name}`;
+                out[locKey] = loc;
+                const offset = loc & 0xffff;
+                const len = (loc >>> 16) & 0xffff;
+                if (offset >= 0 && len > 0 && offset + len <= raw.length) {
+                    const bytes = raw.slice(offset, offset + len);
+                    // 去掉末尾 \0
+                    const nul = bytes.indexOf(0);
+                    const slice = nul >= 0 ? bytes.slice(0, nul) : bytes;
+                    out[field.name] = decoder.decode(slice);
+                }
+                else {
+                    out[field.name] = "";
+                }
+            }
+            continue;
+        }
         const v = readFieldValue(view, field);
         if (v !== undefined) {
             out[key] = v;
@@ -174,25 +203,60 @@ export function parseAllFieldsFromRaw(raw, format) {
     return out;
 }
 /**
- * 将 perf 文本里 raw 段的 hex 行拼成连续字节（小端：每行一个数值，宽度由 hex 位数决定，4→2 字节，8→4 字节，16→8 字节）
+ * 将 perf 文本里 raw 段的 hex 行拼成连续字节（小端）。
+ *
+ * 规则：每行视为一个整数 token（如 0x12345678），按该 token 的字节宽度转成 LE 字节序。
+ * 例如：0x12345678 => [0x78, 0x56, 0x34, 0x12]
  */
-export function rawHexLinesToBuffer(lines) {
+export function rawHexLinesToBuffer(lines, endian = "le") {
     const chunks = [];
     for (const { hex } of lines) {
-        const s = hex.replace(/^0x/i, "").trim();
+        let s = hex.replace(/^0x/i, "").trim();
         if (!s)
             continue;
-        const byteLen = Math.ceil(s.length / 2);
-        const width = byteLen <= 2 ? 2 : byteLen <= 4 ? 4 : 8;
+        if (s.length % 2 === 1)
+            s = "0" + s;
+        const byteLen = s.length / 2;
         let value = BigInt("0x" + s);
-        const mask = (1n << BigInt(width * 8)) - 1n;
-        value &= mask;
-        for (let i = 0; i < width; i++) {
-            chunks.push(Number((value >> BigInt(8 * i)) & 0xffn));
+        if (endian === "le") {
+            for (let i = 0; i < byteLen; i++) {
+                chunks.push(Number((value >> BigInt(8 * i)) & 0xffn));
+            }
+        }
+        else {
+            for (let i = byteLen - 1; i >= 0; i--) {
+                chunks.push(Number((value >> BigInt(8 * i)) & 0xffn));
+            }
         }
     }
     return Uint8Array.from(chunks);
 }
+/**
+ * 将字节缓冲按小端编码为 raw hex 行。
+ * 默认每行 4 字节（即 8 hex digits）。
+ */
+export function bufferToRawHexLines(raw, bytesPerLine = 4, endian = "le") {
+    if (bytesPerLine <= 0)
+        return [];
+    const out = [];
+    for (let i = 0; i < raw.length; i += bytesPerLine) {
+        const end = Math.min(i + bytesPerLine, raw.length);
+        let value = 0n;
+        if (endian === "le") {
+            for (let j = 0; j < end - i; j++) {
+                value |= BigInt(raw[i + j]) << BigInt(8 * j);
+            }
+        }
+        else {
+            for (let j = 0; j < end - i; j++) {
+                value = (value << 8n) | BigInt(raw[i + j]);
+            }
+        }
+        const width = (end - i) * 2;
+        out.push({ hex: "0x" + value.toString(16).padStart(width, "0") });
+    }
+    return out;
+}
 /**
  * 从多个 trace_format 文本构建解析器集合（按 event ID 索引）
  */
@@ -211,6 +275,9 @@ export function buildTraceParserRegistry(formatTexts) {
     return { byEventId, commonFormat };
 }
 function formatArgBySpecifier(value, spec) {
+    if (spec.toLowerCase() === "s") {
+        return typeof value === "string" ? value : String(value);
+    }
     const isBig = typeof value === "bigint";
     const lower = spec.toLowerCase();
     if (lower === "x") {
@@ -225,36 +292,51 @@ function formatArgBySpecifier(value, spec) {
 }
 function renderPrintFmt(printFmt, printArgs, fieldMap) {
     const values = [];
-    for (const expr of printArgs ?? []) {
-        const em = expr.match(/^REC->(\w+)(?:\[(\d+)])?$/);
-        if (!em) {
-            values.push(0);
-            continue;
+    function evalExpr(exprRaw) {
+        // 允许表达式带括号与空格
+        let expr = exprRaw.trim();
+        // 去掉外层括号（可能不止一层，也可能只有一侧被 match 捕获到）
+        while (expr.startsWith("("))
+            expr = expr.slice(1).trim();
+        while (expr.endsWith(")"))
+            expr = expr.slice(0, -1).trim();
+        expr = expr.replace(/\s+/g, "");
+        // (REC->__data_loc_reason&0xffff) / (REC->__data_loc_reason>>16)
+        let m = expr.match(/^REC->(\w+)&0xffff$/);
+        if (m) {
+            const v = fieldMap[m[1]];
+            const n = typeof v === "number" ? v : 0;
+            return n & 0xffff;
         }
-        const name = em[1];
-        const idxRaw = em[2];
+        m = expr.match(/^REC->(\w+)>>16$/);
+        if (m) {
+            const v = fieldMap[m[1]];
+            const n = typeof v === "number" ? v : 0;
+            return (n >>> 16) & 0xffff;
+        }
+        // REC->field 或 REC->field[idx]
+        m = expr.match(/^REC->(\w+)(?:\[(\d+)])?$/);
+        if (!m)
+            return 0;
+        const name = m[1];
+        const idxRaw = m[2];
         const v = fieldMap[name];
         if (idxRaw !== undefined) {
             const idx = parseInt(idxRaw, 10);
             if (Array.isArray(v) && idx >= 0 && idx < v.length) {
-                values.push(v[idx]);
-            }
-            else {
-                values.push(0);
+                return v[idx];
             }
+            return 0;
         }
-        else if (Array.isArray(v)) {
-            values.push(v[0] ?? 0);
-        }
-        else if (v !== undefined) {
-            values.push(v);
-        }
-        else {
-            values.push(0);
-        }
+        if (Array.isArray(v))
+            return v[0] ?? 0;
+        return v ?? 0;
+    }
+    for (const expr of printArgs ?? []) {
+        values.push(evalExpr(expr));
     }
     let valueIdx = 0;
-    return printFmt.replace(/%[0-9]*[lh]*([duxX])/g, (_all, spec) => {
+    return printFmt.replace(/%[0-9]*[lh]*([duxXsS])/g, (_all, spec) => {
         const v = values[valueIdx++] ?? 0;
         return formatArgBySpecifier(v, spec);
     });
@@ -305,22 +387,29 @@ export function decodeRawByRegistry(raw, registry) {
  * - 若找到 common_type 对应解析器：将 print fmt 渲染结果写入 `sample.raw.lines`（一行）。
  * - 若找不到解析器：放弃解析，并将 `common_type` 写入 `sample.raw.lines`（一行）。
  */
-export function decodePerfRawData(perfData, registry) {
+export function decodePerfRawData(perfData, registry, options = {}) {
     return {
         recordSamples: perfData.recordSamples.map((sample) => {
             if (!sample.raw || sample.raw.lines.length === 0)
                 return sample;
             const raw = rawHexLinesToBuffer(sample.raw.lines);
             const decoded = decodeRawByRegistry(raw, registry);
-            const replacement = decoded.renderedText ??
+            const base = decoded.renderedText ??
                 (decoded.commonType !== undefined ? String(decoded.commonType) : "");
-            if (!replacement)
+            if (!base)
                 return sample;
+            const commonLine = options.keepCommonFields
+                ? Object.entries(decoded.commonFields)
+                    .map(([k, v]) => `${k}:${typeof v === "bigint" ? v.toString(10) : v}`)
+                    .join(" ")
+                : "";
             return {
                 ...sample,
                 raw: {
                     ...sample.raw,
-                    lines: [{ hex: replacement }],
+                    lines: options.keepCommonFields && commonLine.length > 0
+                        ? [{ hex: base }, { hex: commonLine }]
+                        : [{ hex: base }],
                 },
             };
         }),

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "hiperf_txt_parser",
-  "version": "1.0.8",
+  "version": "1.0.10",
   "description": "Parse perf data.txt and output structured TypeScript data",
   "type": "module",
   "main": "dist/index.js",