hiperf_txt_parser 1.0.13 → 1.0.15

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, bufferToRawHexLines, buildTraceParserRegistry, decodeRawByRegistry, decodePerfRawData, } from "./traceFormat.js";
-export type { ParsedTraceFormat, TraceFormatField, TraceParserRegistry, DecodedRawSample, DecodePerfRawDataOptions, DecodeRawByRegistryOptions, TracePrintMode, Endian, } from "./traceFormat.js";
+export { parseTraceFormat, parseCommonFieldsFromRaw, parseAllFieldsFromRaw, rawHexLinesToBuffer, bufferToRawHexLines, buildTraceParserRegistry, decodeRawByRegistry, decodePerfRawData, formatPrintfValue, tokenizePrintfFormat, } from "./traceFormat.js";
+export type { ParsedTraceFormat, TraceFormatField, TraceParserRegistry, DecodedRawSample, DecodePerfRawDataOptions, DecodeRawByRegistryOptions, TracePrintMode, PrintfSpec, 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, bufferToRawHexLines, buildTraceParserRegistry, decodeRawByRegistry, decodePerfRawData, } from "./traceFormat.js";
+export { parseTraceFormat, parseCommonFieldsFromRaw, parseAllFieldsFromRaw, rawHexLinesToBuffer, bufferToRawHexLines, buildTraceParserRegistry, decodeRawByRegistry, decodePerfRawData, formatPrintfValue, tokenizePrintfFormat, } from "./traceFormat.js";

package/dist/serializer.d.ts

@@ -3,13 +3,15 @@ import type { PerfData } from "./types.js";
  * 将 PerfData 导出为与 sample/perf_data.txt 相同格式的文本（含缩进与每行前缀）
  */
 export declare function formatPerfDataToText(data: PerfData): string;
-/** 导出用 JSON 的单项：issuce 固定为 "unknow"，call_chain 为 frames 逐行拼接 */
-export interface RecordSampleJsonExportItem {
+/**
+ * 导出用 JSON 单项：固定含 `issuce`、`call_chain`；若存在 `traceFieldDict`，其键值对与二者**同级**平铺。
+ */
+export type RecordSampleJsonExportItem = {
     issuce: "unknow";
     call_chain: string;
-}
+} & Record<string, string>;
 /**
- * 将 PerfData 转为导出用 JSON 结构：顶层为数组，每项为 { issuce: "unknow", call_chain: "" }，
- * call_chain 由对应 record sample 的 callchainFrames.frames 逐行拼接得到，无则为空字符串。
+ * 将 PerfData 转为导出用 JSON 结构：顶层为数组。
+ * call_chain 由 callchainFrames.frames 拼接；有 traceFieldDict 时各字段平铺在对象顶层（与 call_chain 同级）。
  */
 export declare function formatPerfDataToJson(data: PerfData): RecordSampleJsonExportItem[];

package/dist/serializer.js

@@ -19,7 +19,18 @@ function serializeOneSample(sample) {
             lines.push(`    ${addr}`);
         }
     }
-    if (sample.raw) {
+    if (sample.traceFieldDict && Object.keys(sample.traceFieldDict).length > 0 && sample.raw) {
+        lines.push(`  raw size=${sample.raw.size}`);
+        for (const [k, v] of Object.entries(sample.traceFieldDict)) {
+            lines.push(`    ${k}: ${v}`);
+        }
+        const n = Object.keys(sample.traceFieldDict).length;
+        for (let i = n; i < sample.raw.lines.length; i++) {
+            const { hex, short } = sample.raw.lines[i];
+            lines.push(short ? `    ${hex} (${short})` : `    ${hex}`);
+        }
+    }
+    else if (sample.raw) {
         lines.push(`  raw size=${sample.raw.size}`);
         for (const { hex, short } of sample.raw.lines) {
             lines.push(short ? `    ${hex} (${short})` : `    ${hex}`);
@@ -47,12 +58,20 @@ export function formatPerfDataToText(data) {
     return data.recordSamples.map(serializeOneSample).join("\n\n");
 }
 /**
- * 将 PerfData 转为导出用 JSON 结构：顶层为数组，每项为 { issuce: "unknow", call_chain: "" }，
- * call_chain 由对应 record sample 的 callchainFrames.frames 逐行拼接得到，无则为空字符串。
+ * 将 PerfData 转为导出用 JSON 结构：顶层为数组。
+ * call_chain 由 callchainFrames.frames 拼接；有 traceFieldDict 时各字段平铺在对象顶层（与 call_chain 同级）。
  */
 export function formatPerfDataToJson(data) {
-    return data.recordSamples.map((s) => ({
-        issuce: "unknow",
-        call_chain: s.callchainFrames ? s.callchainFrames.frames.join("\n") : "",
-    }));
+    return data.recordSamples.map((s) => {
+        const item = {
+            issuce: "unknow",
+            call_chain: s.callchainFrames ? s.callchainFrames.frames.join("\n") : "",
+        };
+        if (s.traceFieldDict) {
+            for (const [k, v] of Object.entries(s.traceFieldDict)) {
+                item[k] = v;
+            }
+        }
+        return item;
+    });
 }

package/dist/traceFormat.d.ts

@@ -75,6 +75,30 @@ export declare function bufferToRawHexLines(raw: Uint8Array, bytesPerLine?: numb
  * 从多个 trace_format 文本构建解析器集合（按 event ID 索引）
  */
 export declare function buildTraceParserRegistry(formatTexts: string[]): TraceParserRegistry;
+/** 与 C printf 接近的转换说明符（不含 * 动态宽度等） */
+export interface PrintfSpec {
+    flags: string;
+    width?: number;
+    precision?: number;
+    length: string;
+    conv: string;
+}
+/**
+ * 按 C printf 子集格式化：标志 -+ #0 空格，宽度，精度，长度 hl ll hh，转换 d i u x X o s p。
+ */
+export declare function formatPrintfValue(value: number | bigint | string, spec: PrintfSpec): string;
+/**
+ * 将 print fmt 拆成文本、`%%`、与带参数的转换说明符。
+ */
+export declare function tokenizePrintfFormat(printFmt: string): Array<{
+    kind: "text";
+    text: string;
+} | {
+    kind: "literal_pct";
+} | {
+    kind: "spec";
+    spec: PrintfSpec;
+}>;
 /**
  * 对一条 raw 数据先解 common_*，再按 common_type 选择解析器，最后按 print fmt 渲染。
  * 找不到解析器时返回 skipped=true，并打印 common_type。
@@ -88,6 +112,7 @@ export declare function decodeRawByRegistry(raw: Uint8Array, registry: TracePars
  *
  * 规则：
  * - 若找到 common_type 对应解析器：将 print fmt 渲染结果写入 `sample.raw.lines`（一行）。
+ * - fieldDict：写入 `traceFieldDict`，`raw.lines` 为每行一条 `key: value`；`raw.size` 为行数（含可选 common 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, options?: DecodePerfRawDataOptions): PerfData;

package/dist/traceFormat.js

@@ -274,22 +274,232 @@ export function buildTraceParserRegistry(formatTexts) {
     }
     return { byEventId, commonFormat };
 }
-const PRINT_FMT_SPEC_RE = /%[0-9]*[lh]*([duxXsS])/g;
-function formatArgBySpecifier(value, spec) {
-    if (spec.toLowerCase() === "s") {
-        return typeof value === "string" ? value : String(value);
+const PRINT_FLAG_CHARS = new Set(["-", "+", "#", "0", " "]);
+const VALID_PRINTF_CONV = new Set(["d", "i", "u", "x", "X", "o", "s", "S", "p", "P"]);
+function hasPrintfFlag(flags, f) {
+    return flags.includes(f);
+}
+function toBigIntForPrintf(value) {
+    if (typeof value === "bigint")
+        return value;
+    if (typeof value === "string")
+        return 0n;
+    if (!Number.isFinite(value))
+        return 0n;
+    return BigInt(Math.trunc(value));
+}
+/** 按 64 位无符号解释（与内核 trace 常见 long 一致） */
+function asUInt64(n) {
+    return n & ((1n << 64n) - 1n);
+}
+function applyWidth(s, width, leftAlign, zeroPad) {
+    if (width === undefined || s.length >= width)
+        return s;
+    const padLen = width - s.length;
+    const ch = zeroPad && !leftAlign ? "0" : " ";
+    if (leftAlign)
+        return s + ch.repeat(padLen);
+    return ch.repeat(padLen) + s;
+}
+/** 带符号十进制：支持 precision、0 宽度、+- 空格 */
+function formatPrintfD(n, spec) {
+    const neg = n < 0n;
+    const abs = neg ? -n : n;
+    let digits = abs.toString(10);
+    const prec = spec.precision !== undefined ? spec.precision : 1;
+    digits = digits.padStart(Math.max(prec, digits.length), "0");
+    let sign = "";
+    if (neg)
+        sign = "-";
+    else if (hasPrintfFlag(spec.flags, "+"))
+        sign = "+";
+    else if (hasPrintfFlag(spec.flags, " "))
+        sign = " ";
+    const body = sign + digits;
+    if (spec.width === undefined || body.length >= spec.width)
+        return body;
+    const padLen = spec.width - body.length;
+    if (hasPrintfFlag(spec.flags, "0") && !hasPrintfFlag(spec.flags, "-") && spec.precision === undefined) {
+        if (sign)
+            return sign + "0".repeat(padLen) + digits;
+        return "0".repeat(padLen) + body;
+    }
+    if (hasPrintfFlag(spec.flags, "-"))
+        return body + " ".repeat(padLen);
+    return " ".repeat(padLen) + body;
+}
+function formatPrintfUnsignedRadix(unsigned, spec, radix, upper) {
+    let body = radix === 16
+        ? upper
+            ? unsigned.toString(16).toUpperCase()
+            : unsigned.toString(16)
+        : unsigned.toString(radix);
+    const precDefault = 1;
+    const prec = spec.precision !== undefined ? spec.precision : precDefault;
+    body = body.padStart(Math.max(prec, body.length), "0");
+    if (hasPrintfFlag(spec.flags, "#")) {
+        if (radix === 16)
+            body = (upper ? "0X" : "0x") + body;
+        else if (radix === 8 && body !== "0" && !body.startsWith("0"))
+            body = "0" + body;
+    }
+    const zeroPad = hasPrintfFlag(spec.flags, "0") &&
+        !hasPrintfFlag(spec.flags, "-") &&
+        spec.precision === undefined;
+    return applyWidth(body, spec.width, hasPrintfFlag(spec.flags, "-"), zeroPad);
+}
+/**
+ * 按 C printf 子集格式化：标志 -+ #0 空格，宽度，精度，长度 hl ll hh，转换 d i u x X o s p。
+ */
+export function formatPrintfValue(value, spec) {
+    const conv = spec.conv;
+    const cl = conv.toLowerCase();
+    if (conv === "s" || conv === "S") {
+        let s = typeof value === "string" ? value : String(value);
+        if (spec.precision !== undefined) {
+            s = s.slice(0, spec.precision);
+        }
+        return applyWidth(s, spec.width, hasPrintfFlag(spec.flags, "-"), false);
     }
-    const isBig = typeof value === "bigint";
-    const lower = spec.toLowerCase();
-    if (lower === "x") {
-        return isBig ? value.toString(16) : Math.trunc(value).toString(16);
+    const n = toBigIntForPrintf(value);
+    if (conv === "p" || conv === "P") {
+        const u = asUInt64(n);
+        const upper = conv === "P";
+        let body = upper ? u.toString(16).toUpperCase() : u.toString(16);
+        const prec = spec.precision !== undefined ? spec.precision : 1;
+        body = body.padStart(Math.max(prec, body.length), "0");
+        const zeroPad = hasPrintfFlag(spec.flags, "0") &&
+            !hasPrintfFlag(spec.flags, "-") &&
+            spec.precision === undefined;
+        return applyWidth(body, spec.width, hasPrintfFlag(spec.flags, "-"), zeroPad);
     }
-    if (lower === "u") {
-        if (isBig)
-            return (value < 0n ? 0n : value).toString(10);
-        return Math.max(0, Math.trunc(value)).toString(10);
+    if (cl === "d" || cl === "i") {
+        return formatPrintfD(n, spec);
+    }
+    if (cl === "u" || cl === "x" || cl === "X" || cl === "o") {
+        const u = asUInt64(n);
+        if (cl === "u") {
+            let body = u.toString(10);
+            const prec = spec.precision !== undefined ? spec.precision : 1;
+            body = body.padStart(Math.max(prec, body.length), "0");
+            const zeroPad = hasPrintfFlag(spec.flags, "0") &&
+                !hasPrintfFlag(spec.flags, "-") &&
+                spec.precision === undefined;
+            return applyWidth(body, spec.width, hasPrintfFlag(spec.flags, "-"), zeroPad);
+        }
+        if (cl === "o") {
+            return formatPrintfUnsignedRadix(u, spec, 8, false);
+        }
+        return formatPrintfUnsignedRadix(u, spec, 16, conv === "X");
+    }
+    return String(value);
+}
+/**
+ * 将 print fmt 拆成文本、`%%`、与带参数的转换说明符。
+ */
+export function tokenizePrintfFormat(printFmt) {
+    const out = [];
+    let i = 0;
+    let textBuf = "";
+    const flushText = () => {
+        if (textBuf.length > 0) {
+            out.push({ kind: "text", text: textBuf });
+            textBuf = "";
+        }
+    };
+    while (i < printFmt.length) {
+        if (printFmt[i] !== "%") {
+            textBuf += printFmt[i];
+            i++;
+            continue;
+        }
+        flushText();
+        const pctPos = i;
+        if (i + 1 >= printFmt.length) {
+            textBuf += "%";
+            i++;
+            continue;
+        }
+        if (printFmt[i + 1] === "%") {
+            out.push({ kind: "literal_pct" });
+            i += 2;
+            continue;
+        }
+        i++;
+        let flags = "";
+        while (i < printFmt.length && PRINT_FLAG_CHARS.has(printFmt[i])) {
+            flags += printFmt[i];
+            i++;
+        }
+        let width;
+        if (i < printFmt.length && printFmt[i] >= "0" && printFmt[i] <= "9") {
+            let w = 0;
+            while (i < printFmt.length && printFmt[i] >= "0" && printFmt[i] <= "9") {
+                w = w * 10 + (printFmt[i].charCodeAt(0) - 48);
+                i++;
+            }
+            width = w;
+        }
+        let precision;
+        if (i < printFmt.length && printFmt[i] === ".") {
+            i++;
+            if (i < printFmt.length && printFmt[i] >= "0" && printFmt[i] <= "9") {
+                let p = 0;
+                while (i < printFmt.length && printFmt[i] >= "0" && printFmt[i] <= "9") {
+                    p = p * 10 + (printFmt[i].charCodeAt(0) - 48);
+                    i++;
+                }
+                precision = p;
+            }
+            else {
+                precision = 0;
+            }
+        }
+        let length = "";
+        if (i < printFmt.length) {
+            if (printFmt[i] === "h") {
+                length = "h";
+                i++;
+                if (i < printFmt.length && printFmt[i] === "h") {
+                    length = "hh";
+                    i++;
+                }
+            }
+            else if (printFmt[i] === "l") {
+                length = "l";
+                i++;
+                if (i < printFmt.length && printFmt[i] === "l") {
+                    length = "ll";
+                    i++;
+                }
+            }
+            else if (printFmt[i] === "L" ||
+                printFmt[i] === "j" ||
+                printFmt[i] === "z" ||
+                printFmt[i] === "t") {
+                length = printFmt[i];
+                i++;
+            }
+        }
+        if (i >= printFmt.length) {
+            textBuf += printFmt.slice(pctPos, i);
+            break;
+        }
+        const conv = printFmt[i];
+        i++;
+        if (!VALID_PRINTF_CONV.has(conv)) {
+            textBuf += printFmt.slice(pctPos, i);
+            continue;
+        }
+        out.push({ kind: "spec", spec: { flags, width, precision, length, conv } });
     }
-    return isBig ? value.toString(10) : Math.trunc(value).toString(10);
+    flushText();
+    return out;
+}
+function extractPrintfSpecs(printFmt) {
+    return tokenizePrintfFormat(printFmt)
+        .filter((t) => t.kind === "spec")
+        .map((t) => t.spec);
 }
 function normalizePrintExpr(exprRaw) {
     let expr = exprRaw.trim();
@@ -336,28 +546,19 @@ function buildPrintArgValues(printArgs, fieldMap) {
     const values = (printArgs ?? []).map((e) => evalPrintArgExpr(e, fieldMap));
     return { normalizedArgs, values };
 }
-/** 与 renderPrintFmt 中 replace 顺序一致：依次取出每个格式符的类型字母 */
-function extractPrintfSpecChars(printFmt) {
-    const specs = [];
-    printFmt.replace(PRINT_FMT_SPEC_RE, (_all, spec) => {
-        specs.push(spec);
-        return _all;
-    });
-    return specs;
-}
 function formatOnePrintArg(idx, spec, fieldMap, normalizedArgs, values) {
     const v = values[idx] ?? 0;
     const argExpr = normalizedArgs[idx] ?? "";
-    if (spec.toLowerCase() === "s") {
+    if (spec.conv === "s" || spec.conv === "S") {
         const m = argExpr.match(/^REC->__data_loc_(\w+)(?:&0xffff)?(?:>>16)?$/);
         if (m) {
             const s = fieldMap[m[1]];
             if (typeof s === "string") {
-                return s;
+                return formatPrintfValue(s, spec);
             }
         }
     }
-    return formatArgBySpecifier(v, spec);
+    return formatPrintfValue(v, spec);
 }
 function fieldKeyFromNormalizedArg(normalizedArg, index) {
     let k = normalizedArg;
@@ -367,7 +568,7 @@ function fieldKeyFromNormalizedArg(normalizedArg, index) {
 }
 function renderPrintFmtAsFieldDict(printFmt, printArgs, fieldMap) {
     const { normalizedArgs, values } = buildPrintArgValues(printArgs, fieldMap);
-    const specs = extractPrintfSpecChars(printFmt);
+    const specs = extractPrintfSpecs(printFmt);
     const out = {};
     const seen = new Map();
     for (let i = 0; i < specs.length; i++) {
@@ -381,11 +582,21 @@ function renderPrintFmtAsFieldDict(printFmt, printArgs, fieldMap) {
 }
 function renderPrintFmt(printFmt, printArgs, fieldMap) {
     const { normalizedArgs, values } = buildPrintArgValues(printArgs, fieldMap);
-    let valueIdx = 0;
-    return printFmt.replace(PRINT_FMT_SPEC_RE, (_all, spec) => {
-        const idx = valueIdx++;
-        return formatOnePrintArg(idx, spec, fieldMap, normalizedArgs, values);
-    });
+    const tokens = tokenizePrintfFormat(printFmt);
+    let argIdx = 0;
+    let out = "";
+    for (const tok of tokens) {
+        if (tok.kind === "text") {
+            out += tok.text;
+        }
+        else if (tok.kind === "literal_pct") {
+            out += "%";
+        }
+        else {
+            out += formatOnePrintArg(argIdx++, tok.spec, fieldMap, normalizedArgs, values);
+        }
+    }
+    return out;
 }
 /**
  * 对一条 raw 数据先解 common_*，再按 common_type 选择解析器，最后按 print fmt 渲染。
@@ -441,6 +652,7 @@ export function decodeRawByRegistry(raw, registry, options = {}) {
  *
  * 规则：
  * - 若找到 common_type 对应解析器：将 print fmt 渲染结果写入 `sample.raw.lines`（一行）。
+ * - fieldDict：写入 `traceFieldDict`，`raw.lines` 为每行一条 `key: value`；`raw.size` 为行数（含可选 common 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, options = {}) {
@@ -461,8 +673,30 @@ export function decodePerfRawData(perfData, registry, options = {}) {
                     .map(([k, v]) => `${k}:${typeof v === "bigint" ? v.toString(10) : v}`)
                     .join(" ")
                 : "";
+            const dict = decoded.renderedFieldDict;
+            const useFieldDict = options.tracePrintMode === "fieldDict" &&
+                dict !== undefined &&
+                Object.keys(dict).length > 0;
+            if (useFieldDict) {
+                const kvLines = Object.entries(dict).map(([k, v]) => ({
+                    hex: `${k}: ${v}`,
+                }));
+                const lines = options.keepCommonFields && commonLine.length > 0
+                    ? [...kvLines, { hex: commonLine }]
+                    : kvLines;
+                return {
+                    ...sample,
+                    traceFieldDict: dict,
+                    raw: {
+                        ...sample.raw,
+                        size: lines.length,
+                        lines,
+                    },
+                };
+            }
             return {
                 ...sample,
+                traceFieldDict: undefined,
                 raw: {
                     ...sample.raw,
                     lines: options.keepCommonFields && commonLine.length > 0

package/dist/types.d.ts

@@ -39,6 +39,8 @@ export interface RecordSample {
     period: number;
     callchain?: CallchainEntry;
     raw?: RawEntry;
+    /** tracePrintMode 为 fieldDict 且解码成功时：print 各参数键值（用于导出 txt/json） */
+    traceFieldDict?: Record<string, string>;
     server?: ServerEntry;
     callchainFrames?: CallchainFramesEntry;
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "hiperf_txt_parser",
-  "version": "1.0.13",
+  "version": "1.0.15",
   "description": "Parse perf data.txt and output structured TypeScript data",
   "type": "module",
   "main": "dist/index.js",