wallacoloo / gparse Goto Github PK

M32 ("print from sd card file") can be formatted like so, according to http://reprap.org/wiki/G-code#M23:_Select_SD_file and https://github.com/ErikZalm/Marlin/blob/058e4465313191bf769e00a347ceccb4fbd1112c/Marlin/Marlin_main.cpp (line 105):

M32 filename.gco
M32 /path/to/file
M32 !filename.gcode#
M32 S<startpos bytes> !filename#

First off, it becomes obvious that there are a variety of syntaxes. Secondly these are clearly not float parameters. Also, not all formats identify a variable name for the filename. For example, blending the first and last example, M32 S10 S11 is ambiguous - the file could either be named 'S10', and the offset is 11 bytes OR the file could be named 'S11' and the offset is 10 bytes. "Correctly" parsing this ambigious example is literally impossible.

There are a few ways to fix this:

1. require filenames to have an extension (123.456 is now a valid filename, so then all other arguments must be integers. Can also require filenames have at least 1 non-leading letter to separate them from floats).
2. Require paths to be full (start with the a '/', or 'C:...'). This could be useful, because we can control the path formatting returned from M20.
3. Require paths to start with '!' as in example 3&4. This might lead to incompatibilities with (most) hosts.

I vote for #2, as it would be the easiest to implement, and doesn't have any edge-cases.

Some Windows text editors end their lines with "\r\n" instead of just "\n". Need to make sure both line-endings are supported.

Storing each space-separated value into a vector of strings is costly - especially with all that heap allocation. It may be best to just parse the command all in one go and store the float values (eg X**, Y**, Z**), or to preserve the original string and keep a vector of pointers instead (less string allocation).

Command has functions like isG1() and isM105().

Currently, these use inline constants; isG1() is defined as return isOpcode(0x4731u). This 0x4731u is a magic number, but is really the equivalent of "G1" in big-endian. It is equivalent to ('G'<<8) + '1', but precalculated. So ideally, these magic numbers should be replaced with either the above, or something like isOpcode(makeStr('G', '1')).

Basic tests for gcode parsing & serializing. Not sure what testing framework to use for this.