Skip to main content
DEX (Dalvik Executable) is the binary format used by the Android runtime to package compiled Java and Kotlin code for execution on Android devices. Where a desktop JVM loads individual .class files, the Android build toolchain compiles all classes in an application into one or more .dex files, merging constant pools and optimizing cross-class references. Binary Internals includes a DexFile parser that reads the DEX file section by section, resolving string identifiers into human-readable names and exposing the header, ID tables, and class definition list in an interactive tree.
The DEX parser is still in active development. The header section, string/type/proto/field/method ID tables, and class definitions are parsed, but deeper structures such as method code items, debug information, and annotation data may be incomplete or not yet exposed in the tree. Expect changes in future releases.

​
File Extension

.dex — registered by PluginDescriptorDEX which returns ["dex"] from getExtensions().

​
What Is a DEX File?

A DEX file is produced by the d8 (or older dx) compiler from Android’s build toolchain. It bundles all .class files for an application (or a single dex shard) into a single binary that the Android Runtime (ART) or the older Dalvik VM can load directly. The format is described in detail in the Android DEX format specification and on Wikipedia — Dalvik (software). Every valid DEX file begins with an 8-byte magic sequence split across two arrays:
The four-digit version number embedded in the second magic array (035 in the example above) increases monotonically with format revisions. The parser verifies both magic arrays at open time, and also reads and validates the endianness tag (ENDIAN_CONSTANT or REVERSE_ENDIAN_CONSTANT) from the header before any further parsing.

​
Parsed Sections

The DexFile parser processes the DEX file in the order the sections are referenced by the header:
1

Header (header_item)

The header_item immediately follows the 8-byte magic. It contains the checksum, SHA-1 signature, total file size, header size, endianness tag, link section offset/size, map offset, and the size and offset of every ID and definition section described below. All offsets in the file are byte offsets from the start of the file.
2

String Identifiers (string_id_item[])

header.string_ids_size entries, each a 4-byte offset pointing to a string_data_item. The string_data_item stores a MUTF-8 encoded string preceded by its character count encoded as an unsigned LEB128. Binary Internals resolves these lazily via get_string_ids_string(index).
3

Type Identifiers (type_id_item[])

header.type_ids_size entries, each holding a single descriptor_idx field — an index into the string IDs table — giving the type descriptor string (e.g., Ljava/lang/String; or [I). Resolved via get_type_ids_string(index).
4

Prototype Identifiers (proto_id_item[])

header.proto_ids_size entries. Each proto_id_item contains a shorty_idx (short-form descriptor string index), a return_type_idx (type ID index), and a parameters_off offset pointing to a type_list of parameter types. Parameter type lists are resolved and stored in the data map.
5

Field Identifiers (field_id_item[])

header.field_ids_size entries. Each field_id_item combines a class_idx (type ID), a type_idx (field type), and a name_idx (string ID) to fully describe a field reference.
6

Method Identifiers (method_id_item[])

header.method_ids_size entries. Each method_id_item contains a class_idx (type ID), a proto_idx (prototype ID), and a name_idx (string ID) to fully describe a method reference.
7

Class Definitions (class_def_item[])

header.class_defs_size entries. Each class_def_item includes the class type index, access flags, superclass type index, interfaces offset, source file name index, annotations directory offset, class data offset, and static values offset. Offsets for interfaces, annotations, class data, and static values are resolved and stored in the data map for further parsing.

​
Endianness

DEX files carry an explicit endianness tag in the header. The parser reads this tag early in the constructor and creates a PosDataInputStreamDex configured for either little-endian (ENDIAN_CONSTANT = 0x12345678) or big-endian (REVERSE_ENDIAN_CONSTANT = 0x78563412) byte order. All subsequent multi-byte integer reads use this stream.

​
Data Section

Beyond the ID tables and class definitions, many DEX items are stored at arbitrary offsets in a general data section. The parser collects all referenced offsets into a SortedMap<Long, Class<?>> todoData during the main parse pass and then resolves each one into a FileComponent using reflection-based instantiation. The results are stored in DexFile.data.

​
Constant: NO_INDEX

The value NO_INDEX is used in class_def_item and debug_info_item fields to signal that no index is present (equivalent to a null reference).
For the full DEX format specification, see the official Android documentation: Dalvik Executable (DEX) format.