participle in the field!

[drew-512]

Hi friends and @alecthomas, very impressive work! First allow me to say thanks for gifting it to the world. I love Go and projects like these remind of me of why. :)

I'm leading development of an exciting particle physics project that will be presented a number of research institutions including the LHCb in the coming year, and participle seems like a great fit to handle queries from users (particle physicists) into our package. I'm getting the hang of it, but using participle would go way faster if I had help with a starting point for a simple stateless expr language:

    Graph:  <Vtx>[<Edge><Vtx>]*[, ...]
    Edge:   "-" | "~" | "=" | "--" | "~~"
    Vtx:    <VtxID>["^" | "^^" | "^^^" ]
    VtxID:  uint8 (> 0)      

Basically, this is a language to express a graph, such as:

Example string	# Edge Runs
1	1
1-2-3	1
1-2-3, 2-4	2
1^-2=3^	1
1-2, 1-3	2
1=2, 1~~3	2

Some nice-to-haves:

• Support for /* */ style inline comments (that are ignored)
• Instead of commas delimiting edge "runs", they could instead be optional (and a run would be closed by whitespace or EOS). In general, we want whitespace to be insertable anywhere to help with readability and to prevent physicists from getting tripped up on whitespace nuances.

Once parsed, we would then walk each Graph, perhaps something like:

    for run := range g.EdgeRuns() {
        // walk though `run`, extracting alternating VtxID and edge types
    }

We will be unveiling the project in about 2 months and it will be GPL3. We would love to shoutout to participle and publicly acknowledge it. This is a passion project and are not getting paid, so any help would be greatly appreciated.

Hugs,
Drew

[alecthomas]

I'm glad you're enjoying Participle. Have you been through the tutorial? If you do that you should have basically all of the steps necessary to create a parser for this structure, particularly as you already have it in an EBNF-style form.

Basically each of the rules should end up being a struct. The start of the grammar would look something like this:

type Graph struct {
  Spans []*Span `(@@ ("," @@)*)?`
}

type Span struct {
  Start *Vtx `@@`
  Edge []*Edge `@@+`
}

type Edge struct {
  Kind string `@("-" "-"? | "~" "~"? | "="  )`
  End *Vtx `@@`
}

...

The default lexer should be sufficient I believe but will return individual characters for punctuation, never runs of characters. The parser only matches exact tokens. So "-" "-" will work but "--" will not.

[drew-512]

...so I was thinking ASCII since I feel like we're doing this project in the name of science and everything. But then that means, ya know. It's super important this grammar is something physicists can read and type ad hoc easily. shrug

[drew-512]

Oh, and some of those poor damn chars like ≊ are barely readable in almost any font nor can be easily typed (as in fingers).

[drew-512]

p.s. I think my mind is beautifully blown by starting to see the world and parsing and ASTs in a huge new way. Wild to see things this way.

[alecthomas]

p.s. I think my mind is beautifully blown by starting to see the world and parsing and ASTs in a huge new way. Wild to see things this way.

Yes! That was exactly my reaction when I thought of the Participle approach initially :)

[drew-512]

I guess there's no way to get Go to shush on reflect warnings without taking out the whole warning?

struct field tag `@@` not compatible with reflect.StructTag.Get: bad syntax for struct tag pair