-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Showing
1 changed file
with
212 additions
and
1 deletion.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,2 +1,213 @@ | ||
# zlibisc | ||
Various examples of calling zlib | ||
Leveraging code from other languages: zlib | ||
|
||
# Intro | ||
|
||
Recently I reread [this article](https://community.intersystems.com/post/story-support-how-quest-raw-deflate-compressiondecompression-function-leads-node-callout-server) by @Bernd.Mueller. It's about calling DELFATE function from [zlib library](https://zlib.net/). In this article I'll demonstrate several different approaches to callout libraries, we'll build the same functionality (compress function) in several different languages and compare them. | ||
|
||
# NodeJS | ||
|
||
Let's start with NodeJS. I'm taking the code almost directly from Bernd's article, except it does not use files, but rather direct http connection to pass data. For production use, it would be better to pass request as a body and encode both request and response as base64. Still, here's the [code](https://github.com/intersystems-ru/zlibisc/blob/master/node/zlibserver.js): | ||
|
||
``` | ||
//zlibserver.js | ||
const express = require('express'); | ||
const zlib = require('zlib'); | ||
var app = express(); | ||
app.get('/zlibapi/:text', function(req, res) { | ||
res.type('application/json'); | ||
var text=req.params.text; | ||
try { | ||
zlib.deflate(text, (err, buffer) => { | ||
if (!err) { | ||
res.status(200).send(buffer.toString('binary')); | ||
} else { | ||
res.status(500).json( { "error" : err.message}); | ||
// handle error | ||
} | ||
}); | ||
} | ||
catch(err) { | ||
res.status(500).json({ "error" : err.message}); | ||
return; | ||
} | ||
}); | ||
app.listen(3000, function(){ | ||
console.log("zlibserver started"); | ||
}); | ||
``` | ||
|
||
To start it execute in OS bash (assuming `node` and `npm` installed): | ||
``` | ||
cd <repo>\node | ||
npm install | ||
node ./zlibserver.js | ||
``` | ||
|
||
We're running on port 3000, reading input string from request and returning compressed data in response as is. On a Caché side [http request](https://docs.intersystems.com/latest/csp/docbook/DocBook.UI.Page.cls?KEY=GNET_http) is used to interact with this api: | ||
|
||
``` | ||
/// NodeJS implementation | ||
/// do ##class(isc.zlib.Test).node() | ||
ClassMethod node(text As %String = "Hello World", Output response As %String) As %Status | ||
{ | ||
kill response | ||
set req = ##class(%Net.HttpRequest).%New() | ||
set req.Server = "localhost" | ||
set req.Port = 3000 | ||
set req.Location = "/zlibapi/" _ text | ||
set sc = req.Get(,,$$$NO) | ||
quit:$$$ISERR(sc) sc | ||
set response = req.HttpResponse.Data.Read($$$MaxStringLength) | ||
quit sc | ||
} | ||
``` | ||
|
||
Note, that I'm setting the third argument `set sc = req.Get(,,$$$NO)` - `reset` to zero. If you're writing interface to the outside http(s) server it's best to reuse one request object and just modify it as needed to perform new requests. | ||
|
||
# Java | ||
|
||
[Java Gateway](https://docs.intersystems.com/latest/csp/docbook/DocBook.UI.Page.cls?KEY=EJVG) allows calling arbitrary Java code. Coincidently Java has Deflate class which does exactly what we need: | ||
|
||
``` | ||
package isc.zlib; | ||
import java.util.Arrays; | ||
import java.util.zip.Deflater; | ||
public abstract class Java { | ||
public static byte[] compress(String inputString) { | ||
byte[] output = new byte[inputString.length()*3]; | ||
try { | ||
// Encode a String into bytes | ||
byte[] input = inputString.getBytes("UTF-8"); | ||
// Compress the bytes | ||
Deflater compresser = new Deflater(); | ||
compresser.setInput(input); | ||
compresser.finish(); | ||
int compressedDataLength = compresser.deflate(output); | ||
compresser.end(); | ||
output = Arrays.copyOfRange(output, 0, compressedDataLength); | ||
} catch (java.io.UnsupportedEncodingException ex) { | ||
// handle | ||
} | ||
return output; | ||
} | ||
} | ||
``` | ||
|
||
The problem with this implementation is that it returns `byte[]` which becomes a Stream on Caché side. I have tried to return a string, but hadn't been able to found how to form proper binary string from `byte[]`. If you have any ideas please leave a comment. | ||
To run it place jar from [releases page](https://github.com/intersystems-ru/zlibisc/releases) into `<instance>/bin` folder, load ObjectScript code into your instance and execute: | ||
|
||
``` | ||
write $System.Status.GetErrorText(##class(isc.zlib.Utils).createGateway()) | ||
write $System.Status.GetErrorText(##class(isc.zlib.Utils).updateJar()) | ||
``` | ||
|
||
Check `createGateway` method before running the command. Second argument `javaHome` assumes that `JAVA_HOME` environment variable is set. If it does not, specify home of Java 1.8 JRE as a second argument. After installing run this code to get compressed `text`: | ||
|
||
``` | ||
set gateway = ##class(isc.zlib.Utils).connect() | ||
set response = ##class(isc.zlib.Java).compress(gateway, text) | ||
``` | ||
|
||
|
||
# C | ||
|
||
An InterSystems [Callout library](https://docs.intersystems.com/latest/csp/docbook/DocBook.UI.Page.cls?KEY=BGCL_library) is a shared library that contains your custom Callout functions and the enabling code that allows Caché to use them. | ||
|
||
Here's our Callout library: | ||
|
||
``` | ||
#define ZF_DLL | ||
// Ugly Windows hack | ||
#ifndef ulong | ||
typedef unsigned long ulong; | ||
#endif | ||
#include "string.h" | ||
#include "stdio.h" | ||
#include "stdlib.h" | ||
#include "zlib.h" | ||
#include <cdzf.h> | ||
int Compress(char* istream, CACHE_EXSTRP retval) | ||
{ | ||
ulong srcLen = strlen(istream)+1; // +1 for the trailing `\0` | ||
ulong destLen = compressBound(srcLen); // estimate size needed for the buffer | ||
char* ostream = malloc(destLen); | ||
int res = compress(ostream, &destLen, istream, srcLen); | ||
CACHEEXSTRKILL(retval); | ||
if (!CACHEEXSTRNEW(retval,destLen)) {return ZF_FAILURE;} | ||
memcpy(retval->str.ch,ostream,destLen); // copy to retval->str.ch | ||
return ZF_SUCCESS; | ||
} | ||
ZFBEGIN | ||
ZFENTRY("Compress","cJ",Compress) | ||
ZFEND | ||
``` | ||
|
||
To run it place `dll` or `so` files from [releases page](https://github.com/intersystems-ru/zlibisc/releases) into `<instance>/bin` folder. Repository also contains build scripts for Windows and Linux, execute them to build your own version. | ||
Linux prerequisites: `apt install build-essential zlib1g zlib1g-devel` | ||
Windows prerequisites: [WinBuilds](http://win-builds.org/doku.php) | ||
|
||
To interact with callout library execute: | ||
``` | ||
set path = ##class(isc.zlib.Test).getLibPath() //get path to library file | ||
set response = $ZF(-3, path, "Compress", text) // execute function | ||
do $ZF(-3, "") //unload library | ||
``` | ||
|
||
# System | ||
|
||
A little unexpected in an article about callout mechanisms, but Caché also has built-in [Compress](https://docs.intersystems.com/latest/csp/documatic/%25CSP.Documatic.cls?PAGE=CLASS&LIBRARY=%25SYS&CLASSNAME=%25SYSTEM.Util#METHOD_Compress) (and Decompress function). Call it with: `set response = $extract($SYSTEM.Util.Compress(text), 2, *-1)` | ||
Remember that searching the docs or asking the questions here on the community may save you some time. | ||
|
||
# Comparison | ||
|
||
I have run simple tests (1Kb text, 1 000 000 iterations) on Linux and Windows and got these results. | ||
|
||
Windows: | ||
|
||
| Method | Callout | System | Java | Node | | ||
|--------------|---------|--------|---------|----------| | ||
| Time | 22,77 | 33,41 | 152,73 | 622,51 | | ||
| Speed (Kb/s) | 43912 | 29927 | 6547 | 1606 | | ||
| Overhead, % | -/- | 46,73% | 570,75% | 2633,90% | | ||
|
||
Linux: | ||
|
||
| Method | Callout | System | Java | Node | | ||
|--------------|---------|--------|----------|----------| | ||
| Time | 76,3541 | 76,499 | 147,2436 | 953,7311 | | ||
| Speed (Kb/s) | 13097 | 13072 | 6791 | 1049 | | ||
| Overhead, % | -/- | 0,19% | 92,84% | 1149,09% | | ||
|
||
To run tests load code and call: `do ##class(isc.zlib.Test).test(textLength, iterations)` | ||
|
||
|
||
# Conclusion | ||
|
||
With InterSystems products, you can easily leverage existing code in other languages. However, choosing correct implementation is not always easy, you need to take several metrics into account, such as development speed, performance, and maintainability. Do you need to run on different operating systems? Finding answers to these questions can help you decide on the best implementation plan. | ||
|
||
# Links | ||
|
||
- [Repo](https://github.com/intersystems-ru/zlibisc/) | ||
- [Binaries](https://github.com/intersystems-ru/zlibisc/releases) | ||
- [Http requests](https://docs.intersystems.com/latest/csp/docbook/DocBook.UI.Page.cls?KEY=GNET_http) | ||
- [Java Gateway](https://docs.intersystems.com/latest/csp/docbook/DocBook.UI.Page.cls?KEY=EJVG) | ||
- [Callout library](https://docs.intersystems.com/latest/csp/docbook/DocBook.UI.Page.cls?KEY=BGCL_library) | ||
- [Compress function](https://docs.intersystems.com/latest/csp/documatic/%25CSP.Documatic.cls?PAGE=CLASS&LIBRARY=%25SYS&CLASSNAME=%25SYSTEM.Util#METHOD_Compress) |