HDF5Logging.jl

HDF5Logging.hdf5_logger — Function.

hdf5_logger(filename; group_path)

Create a logger that write log messages into the HDF5 file filename within the given group_path (defaults to "log").

Logging

A counter keeps track of an increasing integer index. A log message is written to the given group_path as a group named with this index (converted to a string), with the following fields:

• level::Int, message::String, _module::String, group::String, id::String, file::String, line::String, which are part of every log message
• data, which contains additional key-value pairs as passed on by the user. Keys are strings.

Reading logs

length(logger) returns the last index of logged messages, which can be be accessed with logger[i]. The latter returns a NamedTuple, or nothing for no such message.

Example

julia> using HDF5Logging, Logging

julia> logger = hdf5_logger(tempname())
Logging into HDF5 file /tmp/jl_IbbUvj, 0 messages in “log”

julia> # write log

julia> with_logger(logger) do
       @info "very informative" a = 1
       end

julia> # read log

julia> logger[1]
(level = Info, message = "very informative", _module = "Main", group = "REPL[46]", id = "Main_7a40b9cc", file = "REPL[46]", line = 2, data = ["a" => 1])

Notes

1. The HDF5 file can contain other data, ideally in other groups than group_path.
2. Contiguity of message indexes is not checked. This package will create them in order, starting at 1, but if you delete some with another tool then getindex will just return nothing.
3. A lock is used, so a shared instance should be thread-safe. That said, if you open the same file with another hdf5_logger, consequences are undefined.
4. The HDF5 file is not kept open when not accessed. This is slower, but should help ensure robust operation.