Replace Netty based RFC 1123 date encoding (#3527) by 987Nabil · Pull Request #3550 · zio/zio-http

987Nabil · 2025-06-13T15:03:14Z

netlify · 2025-06-13T15:03:18Z

✅ Deploy Preview for zio-http ready!

Name	Link
🔨 Latest commit	`cac56dd`
🔍 Latest deploy log	https://app.netlify.com/projects/zio-http/deploys/685c1b980c93cd000808ad5d
😎 Deploy Preview	https://deploy-preview-3550--zio-http.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

To edit notification comments on pull requests, go to your Netlify project configuration.

plokhotnyuk · 2025-06-14T09:35:22Z

Here is an efficient implementation of parsing/formatting for different timestamp formats including RFC 1123. Most performance tricks were ported from jsoniter-scala routines for JSON parsing and serialization of different data time values.

Java's DateTimeFormatter is ~10x times slower and affects scalability by eating memory bandwidth greatly. Even when using caching for formatting it introduces much bigger tail latency during refreshing of formatted values.

guizmaii · 2025-06-19T01:03:48Z

+
+  def decodeDate(date: String): Option[ZonedDateTime] = {
+    try {
+      if (date.head.isUpper && date.charAt(3) == ',' && date.charAt(4) == ' ')


date.head throws if date is empty.
Should we maybe prefer to do:

if (date.isEmpty) None else if (date.charAt(0).isUppe && ...) ...

?
Probably a good change for performances

I checked instead, that length is 28 or 29 since only these are valid.
Theoretically 27 is valid (year before 1k) But I think it is okay to ignore

guizmaii · 2025-06-19T01:07:57Z

+  private def decodeRFC1123(date: String): Option[ZonedDateTime] = {
+    {
+      // TODO consider trie-hard approach
+      val c0 = date.head


Prefer date.charAt(0)? It avoids an if (string.isEmpty) check present in the .head method

Related: https://github.com/zio/zio-http/pull/3550/files#r2155802128

987Nabil · 2025-06-20T18:23:32Z

@guizmaii Implemented some benchmarks

Before your suggested changes

Benchmark                                      Mode  Cnt        Score         Error  Units
DateEncodingBenchmark.nettyDecoding           thrpt    3   439288.041 ± 1092690.136  ops/s
DateEncodingBenchmark.nettyDecodingInvalid    thrpt    3   427059.216 ±  895231.682  ops/s
DateEncodingBenchmark.nettyEncoding           thrpt    3   635390.868 ± 1403521.676  ops/s
DateEncodingBenchmark.zioHttpDecoding         thrpt    3  2317595.133 ± 2830576.722  ops/s
DateEncodingBenchmark.zioHttpDecodingInvalid  thrpt    3  2443173.737 ±  219913.014  ops/s
DateEncodingBenchmark.zioHttpEncoding         thrpt    3  1095266.738 ±  378289.638  ops/s

After

Benchmark                                      Mode  Cnt         Score         Error  Units
DateEncodingBenchmark.nettyDecoding           thrpt    3    471698.844 ±  401495.065  ops/s
DateEncodingBenchmark.nettyDecodingInvalid    thrpt    3    433314.955 ±  208199.739  ops/s
DateEncodingBenchmark.nettyEncoding           thrpt    3    596804.828 ± 1649578.829  ops/s
DateEncodingBenchmark.zioHttpDecoding         thrpt    3   2327353.872 ± 2900047.041  ops/s
DateEncodingBenchmark.zioHttpDecodingInvalid  thrpt    3  17922976.451 ± 3374509.274  ops/s
DateEncodingBenchmark.zioHttpEncoding         thrpt    3   1054091.409 ± 1728102.469  ops/s

The invalid case has explicitly strings with invalid length.

nau · 2025-06-22T14:05:40Z

Cool job! Although why not make these changes in Netty code for everyone to benefit?

guizmaii · 2025-06-24T02:10:05Z

+
+  def decodeDate(date: String): Option[ZonedDateTime] = {
+    try {
+      if (date.length != 29 && date.length != 28) return None


@plokhotnyuk Is there any cost of calling date.length twice? Or the compiler and/or JIT will optimise this? 🤔

String length is just a call to an internal field of the string impl. I don't think there is a real cost for this call.

guizmaii · 2025-06-24T02:12:32Z

@nau

Although why not make these changes in Netty code for everyone to benefit?

I can be wrong but I think the idea is probably to have this code in pure Scala so that it compiles to JS and Native and reduces the dependency on Netty

987Nabil · 2025-06-25T13:45:45Z

@nau @guizmaii There are two reasons.

We want to be independent of Netty. Not only for JS/Native, but also to be able to offer different backends in the JVM.
Nettys formatter supports not only RFC1123 dates. We can optimize here by failing for any other date then RFC1123 dates

guizmaii · 2025-06-28T05:11:34Z

+    encodeRFC1123(date)
+
+  def decodeDate(date: String): Option[ZonedDateTime] = {
+    if (date.length != 29 && date.length != 28) return None


@987Nabil
Shouldn't it be || instead of &&? 🤔

987Nabil requested review from jdegoes, kyri-petrou and vigoo as code owners June 13, 2025 15:03

algora-pbc Bot added the 🙋 Bounty claim label Jun 14, 2025

algora-pbc Bot mentioned this pull request Jun 14, 2025

Replace DateEncoding with own impl #3527

Closed

guizmaii reviewed Jun 19, 2025

View reviewed changes

987Nabil added 2 commits June 20, 2025 19:30

Replace Netty based RFC 1123 date encoding (zio#3527)

8650623

Custom RFC1123 encoding for headers (zio#3527)

ed56d66

987Nabil force-pushed the custom-date-encoding branch from 459b391 to d5f8ea8 Compare June 20, 2025 17:31

987Nabil force-pushed the custom-date-encoding branch 2 times, most recently from a7cfa72 to 7196299 Compare June 20, 2025 22:43

He-Pin mentioned this pull request Jun 22, 2025

renderRfc1123DateTimeString can be faster apache/pekko-http#694

Open

guizmaii reviewed Jun 24, 2025

View reviewed changes

Comment thread zio-http/shared/src/main/scala/zio/http/internal/DateEncoding.scala Outdated

guizmaii reviewed Jun 24, 2025

View reviewed changes

Custom RFC1123 decoding for headers (zio#3527)

cac56dd

987Nabil force-pushed the custom-date-encoding branch from 7196299 to cac56dd Compare June 25, 2025 15:53

987Nabil merged commit 782f763 into zio:main Jun 27, 2025
43 checks passed

guizmaii reviewed Jun 28, 2025

View reviewed changes

Uh oh!

Conversation

987Nabil commented Jun 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

netlify Bot commented Jun 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅ Deploy Preview for zio-http ready!

Uh oh!

plokhotnyuk commented Jun 14, 2025

Uh oh!

guizmaii Jun 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

987Nabil Jun 20, 2025

Choose a reason for hiding this comment

Uh oh!

guizmaii Jun 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

987Nabil commented Jun 20, 2025

Uh oh!

nau commented Jun 22, 2025

Uh oh!

Uh oh!

guizmaii Jun 24, 2025

Choose a reason for hiding this comment

Uh oh!

987Nabil Jun 25, 2025

Choose a reason for hiding this comment

Uh oh!

guizmaii commented Jun 24, 2025

Uh oh!

987Nabil commented Jun 25, 2025

Uh oh!

Uh oh!

guizmaii Jun 28, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

987Nabil commented Jun 13, 2025 •

edited

Loading

netlify Bot commented Jun 13, 2025 •

edited

Loading

guizmaii Jun 19, 2025 •

edited

Loading

guizmaii Jun 19, 2025 •

edited

Loading