Decompiling Java and .NET

Press F3 on a compiled file and see source instead of bytes. Two plugins do this — one for Java (.class, .jar, .apk, .dex) and one for .NET (.dll, .exe, .winmd, .netmodule) — and they behave the same way, so this page covers both. Each can be turned off or removed on its own in Configuration ▸ Plugins….

An archive shows as a tree of its classes; a single class shows as one file. Decompile to Sources in the Commands menu writes the result out and puts it in a panel, so you can search, compare and copy from it like any other folder of source.

You install the engine

No decompiler is bundled and nothing is downloaded for you. That is deliberate on two counts: JD-Core, the best-known Java decompiler, is GPLv3 and could not ship inside an Apache-2.0 app — and engines improve, so swapping one should not require a new version of Peach Commander.

Engine Folder… in the viewer opens the folder they belong in. The README there names each engine and its licence.

Java	CFR, Vineflower, Procyon, jadx (for Android .dex and .apk), and javap for plain bytecode
.NET	ILSpy, and monodis for IL

Check Engines runs each engine's version command and tells you three different things apart: installed and working, not installed, and installed but unable to run — a Java tool without a JDK is present and still cannot start, and only actually running it reveals that.

An engine is described by data rather than code, so you can add one yourself:

[cfr]
kinds  = class, jar
tool   = java
args   = -jar {engine} {input}
engine = ~/Library/Application Support/PeachCommander/decompilers/cfr.jar
output = stdout

When more than one engine can handle a file, the first available is used unless you pick one. With two installed, Compare shows both results side by side — useful when one engine gives up on a method the other handles.

Searching inside compiled code

Search all classes looks through the decompiled text rather than the bytes, so you can find a string literal or a method name in a JAR.

Decompiling during a content search over many files is a separate switch, off by default: producing the text can mean running the engine once per class, which on a slow machine is not a reasonable thing to spend a search on. The main search dialog asks separately; this refuses it here too.

Cache and limits

Results are cached, because decompiling the same class twice is pure waiting. The settings hold how many days to keep results and a size limit for the cache; Clear Cache Now empties it and reports how much it freed.

Two timeouts guard against an engine that will not finish: one for a single class or type, one for a whole archive. Both accept 0, which means "use the engine's own default".