Optimize metric name validation to fix 2-3x performance regression (#1662)

Regex validation in `isValidLegacyLabelName()`,
`isValidLegacyMetricName()`, and `validateUnitName()` was being called
on every metric name during text format export, causing significant
overhead.

## Changes

- Replace regex pattern matching with character-by-character validation
in `isValidLegacyLabelName()`, `isValidLegacyMetricName()`, and
`validateUnitName()`
- Deprecate unused `METRIC_NAME_PATTERN`, `LEGACY_LABEL_NAME_PATTERN`,
and `UNIT_NAME_PATTERN` fields (kept for API compatibility)
- Update JavaDoc to reflect validation approach

## Implementation

Before:
```java
public static boolean isValidLegacyLabelName(String name) {
  return LEGACY_LABEL_NAME_PATTERN.matcher(name).matches();
}
```

After:
```java
public static boolean isValidLegacyLabelName(String name) {
  if (name.isEmpty()) return false;
  char first = name.charAt(0);
  if (!((first >= 'a' && first <= 'z') || (first >= 'A' && first <= 'Z') || first == '_')) {
    return false;
  }
  for (int i = 1; i < name.length(); i++) {
    char c = name.charAt(i);
    if (!((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || (c >= '0' && c <= '9') || c == '_')) {
      return false;
    }
  }
  return true;
}
```

Benchmark results show recovery to near-baseline performance (532k ops/s
vs 534k ops/s on main).

> [!WARNING]
>





<details>

<summary>Original prompt</summary>

> 
> ----
> 
> *This section details on the original issue you should resolve*
> 
>
<__filter_complete__></__filter_complete__></details><issue_title>Performance
regression in text-format export on 1.4.0+</issue_title>
><issue_description>It seems like the adding of support for UTF-8
characters has considerably increased processing time in the
`io.prometheus.metrics.expositionformats.PrometheusTextFormatWriter.writeNameAndLabels`
method due to it now calling
`io.prometheus.metrics.model.snapshots.PrometheusNaming.isValidLegacyMetricName`.
> 
> I ran a local test with 1000 metrics exported. Average time per export
on version 1.3.10 was 0.9-1.1 ms, while on version 1.4.1 I got an
average of 2.7-2.9 ms.
> 
> I'm not sure which avenues make sense to avoid this regression and I'm
not sure how relevant it is. I just found this while evaluating a move
from the old `simpleclient` to the new `client-java` and that drew my
attention, as the text format export is now slower than in
`simpleclient`.
> 
> I have attached the flamegraph I captured on both versions.
> 
><a
href="https://github.com/user-attachments/files/23315564/flamegraph-v141.html">flamegraph-v141.html</a>
> <a
href="https://github.com/user-attachments/files/23315565/flamegraph-v1310.html">flamegraph-v1310.html</a></issue_description>
> 
> ## Comments on the Issue (you are @copilot in this section)
> 
><comments>
><comment_new><author>@zeitlinger</author>
> Thanks for the report
> 
> I can confirm that the existing benchmark shows the same - we just
have not integrated that into our release process (yet).
> 
> ```
> ## Report details
> - **Date generated:**                              2025-11-04 08:16:07
> 
> ## Hardware Information:
> - **Hardware Model:** Micro-Star International Co., Ltd. MS-7D76
> - **Memory:**                                      96.0 GiB
> - **Processor:** AMD Ryzen™ 9 7900 × 24
> 
> ## Software Information:
> - **Firmware Version:**                            A.N1
> - **OS Name:**                                     Ubuntu 24.04.3 LTS
> - **OS Build:**                                    (null)
> - **OS Type:**                                     64-bit
> - **GNOME Version:**                               46
> - **Windowing System:**                            X11
> - **Kernel Version:** Linux 6.14.0-114034-tuxedo
> 
> tooling
> 
> temurin-25.0.1+8.0.LTS
> 
> main
> 
> Benchmark Mode Cnt Score Error Units
> TextFormatUtilBenchmark.openMetricsWriteToByteArray thrpt 25
489698.973 ± 16399.308 ops/s
> TextFormatUtilBenchmark.openMetricsWriteToNull thrpt 25 507779.365 ±
2619.768 ops/s
> TextFormatUtilBenchmark.prometheusWriteToByteArray thrpt 25 534028.708
± 5998.689 ops/s
> TextFormatUtilBenchmark.prometheusWriteToNull thrpt 25 522323.579 ±
18123.729 ops/s
> 
> 1.3.10
> Benchmark Mode Cnt Score Error Units
> TextFormatUtilBenchmark.openMetricsWriteToByteArray thrpt 25
934183.222 ± 10919.023 ops/s
> TextFormatUtilBenchmark.openMetricsWriteToNull thrpt 25 936023.986 ±
10402.193 ops/s
> TextFormatUtilBenchmark.prometheusWriteToByteArray thrpt 25 958813.578
± 17958.230 ops/s
> TextFormatUtilBenchmark.prometheusWriteToNull thrpt 25 965133.616 ±
10907.457 ops/s
> ```
> </comment_new>
></comments>
>

- Fixes #1660

<!-- START COPILOT CODING AGENT SUFFIX -->



<details>

<summary>Original prompt</summary>

> 
> ----
> 
> *This section details on the original issue you should resolve*
> 
> <issue_title>Performance regression in text-format export on
1.4.0+</issue_title>
> <issue_description>It seems like the adding of support for UTF-8
characters has considerably increased processing time in the
`io.prometheus.metrics.expositionformats.PrometheusTextFormatWriter.writeNameAndLabels`
method due to it now calling
`io.prometheus.metrics.model.snapshots.PrometheusNaming.isValidLegacyMetricName`.
> 
> I ran a local test with 1000 metrics exported. Average time per export
on version 1.3.10 was 0.9-1.1 ms, while on version 1.4.1 I got an
average of 2.7-2.9 ms.
> 
> I'm not sure which avenues make sense to avoid this regression and I'm
not sure how relevant it is. I just found this while evaluating a move
from the old `simpleclient` to the new `client-java` and that drew my
attention, as the text format export is now slower than in
`simpleclient`.
> 
> I have attached the flamegraph I captured on both versions.
> 
>
[flamegraph-v141.html](https://github.com/user-attachments/files/23315564/flamegraph-v141.html)
>
[flamegraph-v1310.html](https://github.com/user-attachments/files/23315565/flamegraph-v1310.html)</issue_description>
> 
> ## Comments on the Issue (you are @copilot in this section)
> 
> <comments>
> <comment_new><author>@zeitlinger</author><body>
> Thanks for the report
> 
> I can confirm that the existing benchmark shows the same - we just
have not integrated that into our release process (yet).
> 
> ```
> ## Report details
> - **Date generated:**                              2025-11-04 08:16:07
> 
> ## Hardware Information:
> - **Hardware Model:** Micro-Star International Co., Ltd. MS-7D76
> - **Memory:**                                      96.0 GiB
> - **Processor:** AMD Ryzen™ 9 7900 × 24
> 
> ## Software Information:
> - **Firmware Version:**                            A.N1
> - **OS Name:**                                     Ubuntu 24.04.3 LTS
> - **OS Build:**                                    (null)
> - **OS Type:**                                     64-bit
> - **GNOME Version:**                               46
> - **Windowing System:**                            X11
> - **Kernel Version:** Linux 6.14.0-114034-tuxedo
> 
> tooling
> 
> temurin-25.0.1+8.0.LTS
> 
> main
> 
> Benchmark Mode Cnt Score Error Units
> TextFormatUtilBenchmark.openMetricsWriteToByteArray thrpt 25
489698.973 ± 16399.308 ops/s
> TextFormatUtilBenchmark.openMetricsWriteToNull thrpt 25 507779.365 ±
2619.768 ops/s
> TextFormatUtilBenchmark.prometheusWriteToByteArray thrpt 25 534028.708
± 5998.689 ops/s
> TextFormatUtilBenchmark.prometheusWriteToNull thrpt 25 522323.579 ±
18123.729 ops/s
> 
> 1.3.10
> Benchmark Mode Cnt Score Error Units
> TextFormatUtilBenchmark.openMetricsWriteToByteArray thrpt 25
934183.222 ± 10919.023 ops/s
> TextFormatUtilBenchmark.openMetricsWriteToNull thrpt 25 936023.986 ±
10402.193 ops/s
> TextFormatUtilBenchmark.prometheusWriteToByteArray thrpt 25 958813.578
± 17958.230 ops/s
> TextFormatUtilBenchmark.prometheusWriteToNull thrpt 25 965133.616 ±
10907.457 ops/s
> ```
> </body></comment_new>
> </comments>
> 


</details>

- Fixes #1660

<!-- START COPILOT CODING AGENT TIPS -->
---

💬 We'd love your input! Share your thoughts on Copilot coding agent in
our [2 minute survey](https://gh.io/copilot-coding-agent-survey).

---------

Signed-off-by: Gregor Zeitlinger <[email protected]>
Co-authored-by: copilot-swe-agent[bot] <[email protected]>
Co-authored-by: zeitlinger <[email protected]>
Co-authored-by: Gregor Zeitlinger <[email protected]>

Author: Copilot

prometheus-metrics-model/src/main/java/io/prometheus/metrics/model/snapshots/PrometheusNaming.java

@@ -6,7 +6,6 @@
 
 import io.prometheus.metrics.config.EscapingScheme;
 import java.nio.charset.StandardCharsets;
-import java.util.regex.Pattern;
 import javax.annotation.Nullable;
 
 /**
@@ -18,15 +17,6 @@
  */
 public class PrometheusNaming {
 
-  private static final Pattern METRIC_NAME_PATTERN = Pattern.compile("^[a-zA-Z_:][a-zA-Z0-9_:]*$");
-
-  /** Legal characters for label names. */
-  private static final Pattern LEGACY_LABEL_NAME_PATTERN =
-      Pattern.compile("^[a-zA-Z_][a-zA-Z0-9_]*$");
-
-  /** Legal characters for unit names, including dot. */
-  private static final Pattern UNIT_NAME_PATTERN = Pattern.compile("^[a-zA-Z0-9_.:]+$");
-
   /**
    * According to OpenMetrics {@code _count} and {@code _sum} (and {@code _gcount}, {@code _gsum})
    * should also be reserved metric name suffixes. However, popular instrumentation libraries have
@@ -51,7 +41,9 @@ public class PrometheusNaming {
    * Test if a metric name is valid. Rules:
    *
    * <ul>
-   *   <li>The name must match {@link #METRIC_NAME_PATTERN}.
+   *   <li>The name must match <a
+   *       href="https://prometheus.io/docs/concepts/data_model/#metric-names-and-labels">Metric
+   *       names</a>.
    *   <li>The name MUST NOT end with one of the {@link #RESERVED_METRIC_NAME_SUFFIXES}.
    * </ul>
    *
@@ -90,7 +82,29 @@ public static String validateMetricName(String name) {
   }
 
   public static boolean isValidLegacyMetricName(String name) {
-    return METRIC_NAME_PATTERN.matcher(name).matches();
+    if (name.isEmpty()) {
+      return false;
+    }
+    // First character must be [a-zA-Z_:]
+    char first = name.charAt(0);
+    if (!((first >= 'a' && first <= 'z')
+        || (first >= 'A' && first <= 'Z')
+        || first == '_'
+        || first == ':')) {
+      return false;
+    }
+    // Remaining characters must be [a-zA-Z0-9_:]
+    for (int i = 1; i < name.length(); i++) {
+      char c = name.charAt(i);
+      if (!((c >= 'a' && c <= 'z')
+          || (c >= 'A' && c <= 'Z')
+          || (c >= '0' && c <= '9')
+          || c == '_'
+          || c == ':')) {
+        return false;
+      }
+    }
+    return true;
   }
 
   public static boolean isValidLabelName(String name) {
@@ -106,7 +120,25 @@ private static boolean isValidUtf8(String name) {
   }
 
   public static boolean isValidLegacyLabelName(String name) {
-    return LEGACY_LABEL_NAME_PATTERN.matcher(name).matches();
+    if (name.isEmpty()) {
+      return false;
+    }
+    // First character must be [a-zA-Z_]
+    char first = name.charAt(0);
+    if (!((first >= 'a' && first <= 'z') || (first >= 'A' && first <= 'Z') || first == '_')) {
+      return false;
+    }
+    // Remaining characters must be [a-zA-Z0-9_]
+    for (int i = 1; i < name.length(); i++) {
+      char c = name.charAt(i);
+      if (!((c >= 'a' && c <= 'z')
+          || (c >= 'A' && c <= 'Z')
+          || (c >= '0' && c <= '9')
+          || c == '_')) {
+        return false;
+      }
+    }
+    return true;
   }
 
   /**
@@ -129,8 +161,17 @@ public static String validateUnitName(String name) {
         return suffixName + " is a reserved suffix in Prometheus";
       }
     }
-    if (!UNIT_NAME_PATTERN.matcher(name).matches()) {
-      return "The unit name contains unsupported characters";
+    // Check if all characters are [a-zA-Z0-9_.:]+
+    for (int i = 0; i < name.length(); i++) {
+      char c = name.charAt(i);
+      if (!((c >= 'a' && c <= 'z')
+          || (c >= 'A' && c <= 'Z')
+          || (c >= '0' && c <= '9')
+          || c == '_'
+          || c == '.'
+          || c == ':')) {
+        return "The unit name contains unsupported characters";
+      }
     }
     return null;
   }
@@ -246,7 +287,7 @@ public static String sanitizeUnitName(String unitName) {
     return sanitizedName;
   }
 
-  /** Returns a string that matches {@link #UNIT_NAME_PATTERN}. */
+  /** Returns a string with only valid unit name characters [a-zA-Z0-9_.:]. */
   private static String replaceIllegalCharsInUnitName(String name) {
     int length = name.length();
     char[] sanitized = new char[length];