diff --git a/ezlog-core/src/decode.rs b/ezlog-core/src/decode.rs
index 04ae4bd..bf8d53a 100644
--- a/ezlog-core/src/decode.rs
+++ b/ezlog-core/src/decode.rs
@@ -3,6 +3,8 @@ use std::{
         self,
         BufRead,
         Cursor,
+        Read,
+        Seek,
         Write,
     },
     str::FromStr,
@@ -16,8 +18,6 @@ use byteorder::{
 };
 use integer_encoding::VarIntReader;
 use log::error;
-use once_cell::sync::OnceCell;
-use regex::Regex;
 use time::{
     format_description::well_known::Rfc3339,
     OffsetDateTime,
@@ -37,68 +37,68 @@ use crate::{
 };
 
 pub fn decode_record(vec: &[u8]) -> Result<crate::EZRecord> {
-    static RE: OnceCell<std::result::Result<Regex, regex::Error>> = OnceCell::new();
-    let regex = RE.get_or_init(|| Regex::new(r"\[(.*?)\]"));
-    let record_str =
-        String::from_utf8(vec.to_vec()).map_err(|e| LogError::Parse(format!("{}", e)))?;
     let mut record_builder = EZRecord::builder();
-    if let Ok(regex) = regex {
-        // Search for the first match
-        if let Some(caps) = regex.captures(&record_str) {
-            if let Some(matched) = caps.get(1) {
-                let header = matched.as_str();
-                let split: Vec<&str> = header.split_whitespace().collect();
-
-                let time = split
-                    .first()
-                    .map(|x| {
-                        OffsetDateTime::parse(x, &Rfc3339).unwrap_or(OffsetDateTime::now_utc())
-                    })
-                    .unwrap_or(OffsetDateTime::now_utc());
-                record_builder.time(time);
-
-                let level = split
-                    .get(1)
-                    .and_then(|x| Level::from_str(x).ok())
-                    .unwrap_or(Level::Trace);
-                record_builder.level(level);
-
-                let target = split.get(2).unwrap_or(&"").to_string();
-                record_builder.target(target);
-
-                let thread_str = split.get(3).unwrap_or(&"");
-                let mut thread_info: Vec<&str> = thread_str.split(':').collect();
-                if !thread_info.is_empty() {
-                    let thread_id = thread_info.last().unwrap_or(&"").to_string();
-                    record_builder.thread_id(thread_id.parse::<usize>().unwrap_or(0));
-                    thread_info.pop();
-                    let thread_name = thread_info.join(":");
-                    record_builder.thread_name(thread_name);
-                }
+    let mut cursor = Cursor::new(vec);
+    let mut buf = vec![];
+    cursor
+        .read_until(b'[', &mut buf)
+        .map_err(|_| LogError::Parse("not found [".to_string()))?;
+    buf.clear();
+    cursor
+        .read_until(b']', &mut buf)
+        .map_err(|_| LogError::Parse("not found ]".to_string()))?;
+    buf.pop();
+
+    let header_buf = String::from_utf8_lossy(&buf);
+    let mut header_split = header_buf.split_whitespace();
+    let time = header_split
+        .next()
+        .map(|x| OffsetDateTime::parse(x, &Rfc3339).unwrap_or(OffsetDateTime::now_utc()))
+        .unwrap_or(OffsetDateTime::now_utc());
+    record_builder.time(time);
+
+    let level = header_split
+        .next()
+        .and_then(|x| Level::from_str(x).ok())
+        .unwrap_or(Level::Trace);
+    record_builder.level(level);
+
+    let target = header_split.next().unwrap_or("");
+    record_builder.target(target.to_string());
+
+    if let Some(thread_str) = header_split.next() {
+        let mut thread_info: Vec<&str> = thread_str.split(':').collect();
+        if !thread_info.is_empty() {
+            let thread_id = thread_info.pop().unwrap_or("");
+            let thread_name: String = thread_info.join(":");
+            record_builder.thread_id(thread_id.parse::<usize>().unwrap_or(0));
+            record_builder.thread_name(thread_name);
+        }
+    }
 
-                #[cfg(feature = "log")]
-                {
-                    if split.len() == 5 {
-                        let file_info: Vec<&str> = split.last().unwrap_or(&"").split(':').collect();
-                        if !file_info.is_empty() {
-                            let file_name = file_info.first().unwrap_or(&"");
-                            let line = file_info.get(1).unwrap_or(&"");
-
-                            record_builder.file(file_name);
-                            record_builder.line(line.parse::<u32>().unwrap_or(0));
-                        }
-                    };
-                }
-                // header not contains square brackets
-                // two square brackets and one space
-                if record_str.len() > header.len() + 3 {
-                    let content = &record_str[header.len() + 3..];
-                    record_builder.content(content.to_string());
-                }
+    #[cfg(feature = "log")]
+    {
+        let file_str = header_split.next().unwrap_or("");
+        if !file_str.is_empty() {
+            let file_info: Vec<&str> = file_str.split(':').collect();
+            if !file_info.is_empty() {
+                let file_name = file_info.first().unwrap_or(&"");
+                let line = file_info.get(1).unwrap_or(&"");
+
+                record_builder.file(file_name);
+                record_builder.line(line.parse::<u32>().unwrap_or(0));
             }
         }
-    }
+    };
 
+    let mut content_buf = vec![];
+    // skip whitespace
+    cursor.seek(io::SeekFrom::Current(1)).unwrap_or(0);
+    cursor
+        .read_to_end(&mut content_buf)
+        .map_err(|e| LogError::Parse(format!("parse content: {}", e)))?;
+    let content = String::from_utf8_lossy(&content_buf);
+    record_builder.content(content.to_string());
     Ok(record_builder.build())
 }