Memory-managed wrapper type.

A convenience alias for Nothing :: Maybe Converter.

Type class for types which can be safely cast to Converter, for instance with toConverter.

Cast to Converter, for types for which this is known to be safe. For general casts, use castTo.

This is the main operation used when converting data. It is to be called multiple times in a loop, and each time it will do some work, i.e. producing some output (in outbuf) or consuming some input (from inbuf) or both. If its not possible to do any work an error is returned.

Note that a single call may not consume all input (or any input at all). Also a call may produce output even if given no input, due to state stored in the converter producing output.

If any data was either produced or consumed, and then an error happens, then only the successful conversion is reported and the error is returned on the next call.

A full conversion loop involves calling this method repeatedly, each time giving it new input and space output space. When there is no more input data after the data in inbuf, the flag ConverterFlagsInputAtEnd must be set. The loop will be (unless some error happens) returning ConverterResultConverted each time until all data is consumed and all output is produced, then ConverterResultFinished is returned instead. Note, that ConverterResultFinished may be returned even if ConverterFlagsInputAtEnd is not set, for instance in a decompression converter where the end of data is detectable from the data (and there might even be other data after the end of the compressed data).

When some data has successfully been converted bytesRead and is set to the number of bytes read from inbuf, and bytesWritten is set to indicate how many bytes was written to outbuf. If there are more data to output or consume (i.e. unless the ConverterFlagsInputAtEnd is specified) then ConverterResultConverted is returned, and if no more data is to be output then ConverterResultFinished is returned.

On error ConverterResultError is returned and error is set accordingly. Some errors need special handling:

IOErrorEnumNoSpace is returned if there is not enough space to write the resulting converted data, the application should call the function again with a larger outbuf to continue.

IOErrorEnumPartialInput is returned if there is not enough input to fully determine what the conversion should produce, and the ConverterFlagsInputAtEnd flag is not set. This happens for example with an incomplete multibyte sequence when converting text, or when a regexp matches up to the end of the input (and may match further input). It may also happen when inbufSize is zero and there is no more data to produce.

When this happens the application should read more input and then call the function again. If further input shows that there is no more data call the function again with the same data but with the ConverterFlagsInputAtEnd flag set. This may cause the conversion to finish as e.g. in the regexp match case (or, to fail again with IOErrorEnumPartialInput in e.g. a charset conversion where the input is actually partial).

After converterConvert has returned ConverterResultFinished the converter object is in an invalid state where its not allowed to call converterConvert anymore. At this time you can only free the object or call converterReset to reset it to the initial state.

If the flag ConverterFlagsFlush is set then conversion is modified to try to write out all internal state to the output. The application has to call the function multiple times with the flag set, and when the available input has been consumed and all internal state has been produced then ConverterResultFlushed (or ConverterResultFinished if really at the end) is returned instead of ConverterResultConverted. This is somewhat similar to what happens at the end of the input stream, but done in the middle of the data.

This has different meanings for different conversions. For instance in a compression converter it would mean that we flush all the compression state into output such that if you uncompress the compressed data you get back all the input data. Doing this may make the final file larger due to padding though. Another example is a regexp conversion, where if you at the end of the flushed data have a match, but there is also a potential longer match. In the non-flushed case we would ask for more input, but when flushing we treat this as the end of input and do the match.

Flushing is not always possible (like if a charset converter flushes at a partial multibyte sequence). Converters are supposed to try to produce as much output as possible and then return an error (typically IOErrorEnumPartialInput).

Since: 2.24

Resets all internal state in the converter, making it behave as if it was just created. If the converter has any internal state that would produce output then that output is lost.

Since: 2.24