Skip Keys¶
The skip functionality in Glaze allows you to conditionally skip struct members during JSON serialization and parsing at compile time. This feature operates at compile time and can differentiate between serialize and parse operations.
Overview¶
By default, Glaze serializes all public C++ struct members to JSON. However, you may need to omit certain fields from your JSON output or input processing.
Meta Context¶
The skip function receives a meta_context parameter that provides compile-time information about the current operation:
namespace glz {
enum struct operation {
serialize, // Writing/serializing to JSON
parse // Reading/parsing from JSON
};
struct meta_context {
operation op = operation::serialize;
};
}
This allows you to implement different skipping logic for serialization versus parsing operations.
Basic Usage¶
To use skip, specialize the glz::meta template for your struct and implement a skip function:
struct my_struct_t {
std::string public_member{};
std::string internal_member{};
};
template <>
struct glz::meta<my_struct_t> {
static constexpr bool skip(const std::string_view key, const meta_context&) {
// Return true to skip the key, false to include it
return key == "internal_member";
}
};
Example 1: Skipping Specific Keys¶
Omit specific member names from the JSON output:
struct skipped_t {
std::string name{};
int age{};
std::string secret_info{};
};
template <>
struct glz::meta<skipped_t> {
static constexpr bool skip(const std::string_view key, const meta_context&) {
if (key == "secret_info") {
return true; // Skip this field
}
return false; // Include other fields
}
};
Usage:
skipped_t obj{"John Doe", 30, "My Top Secret"};
std::string buffer{};
// Writing JSON
glz::write_json(obj, buffer);
// Output: {"name":"John Doe","age":30}
Example 2: Skipping Based on Prefix/Suffix¶
Apply systematic skipping based on key patterns using compile-time string manipulation:
struct prefixed_skipped_t {
std::string user_id{};
std::string user_name{};
std::string temp_data{};
};
template <>
struct glz::meta<prefixed_skipped_t> {
static constexpr bool skip(const std::string_view key, const meta_context&) {
return key.starts_with("temp_"); // Skip keys starting with "temp_"
}
};
Usage:
prefixed_skipped_t obj{"123", "Alice", "temporary value"};
std::string buffer{};
// Writing JSON
glz::write_json(obj, buffer);
// Output: {"user_id":"123","user_name":"Alice"}
Example 3: Operation-Specific Skipping¶
Use the meta_context to implement different skipping behavior for serialization versus parsing:
struct versioned_data_t {
std::string name{};
int version{};
std::string computed_field{};
std::string input_only_field{};
};
template <>
struct glz::meta<versioned_data_t> {
static constexpr bool skip(const std::string_view key, const meta_context& ctx) {
// Skip computed_field during parsing (reading) - it should only be written
if (key == "computed_field" && ctx.op == glz::operation::parse) {
return true;
}
// Skip input_only_field during serialization (writing) - it should only be read
if (key == "input_only_field" && ctx.op == glz::operation::serialize) {
return true;
}
return false;
}
};
Usage:
versioned_data_t obj{"TestData", 1, "computed_value", "input_value"};
std::string buffer{};
// Writing JSON - skips input_only_field
glz::write_json(obj, buffer);
// Output: {"name":"TestData","version":1,"computed_field":"computed_value"}
// Reading JSON - skips computed_field
const char* json = R"({"name":"NewData","version":2,"computed_field":"ignored","input_only_field":"new_input"})";
glz::read_json(obj, json);
// obj.computed_field retains "computed_value", obj.input_only_field becomes "new_input"
Value-Based Skipping with skip_if¶
While skip allows skipping fields based on their key names at compile time, skip_if enables runtime skipping based on the actual field values. This is useful for omitting fields with default values or applying conditional logic based on the data.
Overview¶
The skip_if function is a template method that receives:
- The field value (with type information preserved)
- The key name
- The meta_context (operation type)
This allows you to inspect the actual value and decide whether to skip serialization.
Basic Usage¶
struct my_struct_t {
std::string name = "default";
int age = 0;
std::string city{};
};
template <>
struct glz::meta<my_struct_t> {
template <class T>
static constexpr bool skip_if(T&& value, std::string_view key, const glz::meta_context&) {
using V = std::decay_t<T>;
if constexpr (std::same_as<V, std::string>) {
return key == "name" && value == "default";
}
else if constexpr (std::same_as<V, int>) {
return key == "age" && value == 0;
}
return false;
}
};
Example: Skipping Default Values¶
A common use case is omitting fields that contain their default values to reduce JSON size:
struct user_settings_t {
std::string theme = "light";
int volume = 50;
bool notifications = true;
std::string custom_path{};
};
template <>
struct glz::meta<user_settings_t> {
template <class T>
static constexpr bool skip_if(T&& value, std::string_view key, const glz::meta_context&) {
using V = std::decay_t<T>;
if constexpr (std::same_as<V, std::string>) {
if (key == "theme") return value == "light";
if (key == "custom_path") return value.empty();
}
else if constexpr (std::same_as<V, int>) {
if (key == "volume") return value == 50;
}
else if constexpr (std::same_as<V, bool>) {
if (key == "notifications") return value == true;
}
return false;
}
};
Usage:
user_settings_t settings1{"light", 50, true, ""};
std::string buffer1{};
glz::write_json(settings1, buffer1);
// Output: {} (all fields have default values)
user_settings_t settings2{"dark", 75, false, "/custom/path"};
std::string buffer2{};
glz::write_json(settings2, buffer2);
// Output: {"theme":"dark","volume":75,"notifications":false,"custom_path":"/custom/path"}
user_settings_t settings3{"light", 80, true, ""};
std::string buffer3{};
glz::write_json(settings3, buffer3);
// Output: {"volume":80} (only non-default field included)
Important Notes¶
skip_ifis evaluated at runtime during serialization, unlikeskipwhich is compile-time- The template method allows type-safe inspection of field values using
if constexpr skip_ifis primarily for serialization (writing); during parsing, all present fields are read- You can use both
skipandskip_iftogether:skipis checked first (compile-time), and if a field isn't skipped,skip_ifis then evaluated (runtime) - Use
std::decay_tto get the underlying type without references and cv-qualifiers
Combining skip and skip_if¶
You can use both methods together for maximum flexibility. skip is evaluated first (at compile-time), and if it returns false, skip_if is then evaluated (at runtime):
struct api_response_t {
std::string id{};
std::string secret_key{}; // Never serialize
std::string name{};
int count = 0;
};
template <>
struct glz::meta<api_response_t> {
// Compile-time skip: always exclude secret_key
static constexpr bool skip(const std::string_view key, const glz::meta_context&) {
return key == "secret_key";
}
// Runtime skip: exclude count when it's 0
template <class T>
static constexpr bool skip_if(T&& value, std::string_view key, const glz::meta_context&) {
using V = std::decay_t<T>;
if constexpr (std::same_as<V, int>) {
return key == "count" && value == 0;
}
return false;
}
};
Usage:
api_response_t resp1{"123", "secret", "Widget", 0};
std::string buffer1{};
glz::write_json(resp1, buffer1);
// Output: {"id":"123","name":"Widget"}
// secret_key skipped by skip(), count skipped by skip_if()
api_response_t resp2{"456", "secret", "Gadget", 5};
std::string buffer2{};
glz::write_json(resp2, buffer2);
// Output: {"id":"456","name":"Gadget","count":5}
// secret_key skipped by skip(), count included (non-zero)
Comparison: skip vs skip_if¶
| Feature | skip (Key-Based) |
skip_if (Value-Based) |
|---|---|---|
| Decision time | Compile-time | Runtime |
| Input | Key name only | Key name + field value |
| Use case | Always skip certain fields | Skip based on field values |
| Performance | Zero runtime overhead | Minimal runtime check per field |
| Template | Not templated | Template method |
| Combinable | Can be used with skip_if |
Can be used with skip |
Choose skip when you want to permanently exclude fields, and skip_if when the decision depends on the data. Use both together for maximum control.