buggyorg / buggy-cli Goto Github PK

Currently a mux can only detect recursive nodes inside of one compound. But if the mux is somehow encapsulated it should also work.

Currently versions are quite arbitrary and introduce a lot of extra work. The version should always indicate what is compatible. This mainly concerns the implementations in the component-library.

I would prefer a system where, as long as buggy is under heavy development, that the version stays below 1.0.0 and the minor version indicates the what dependencies are compatible. E.g. for the upcoming v0.6.0 release every node in the component library must have a 0.6.* version. Other version (like 0.5.*) are not valid.

This has the advantage that all version except the current release can be dropped and we use the newest compatible version in this sub release. In clojure files one could say (use 0.6.*) at the beginning of the file to choose compatible implementations.

A disadvantage is, that it is not possible to independently introduce changes to nodes that are not backwards compatible. Those changes would have to wait till the next minor release.

Currently we use the order of inputs / outputs in some places (e.g. partials and code generation). But the inputs are stored as an object and thus have no guaranteed order.

The function sequenceDependencies got rather confusing and requires a cleaner implementation. Moreover the method can currently only process two levels of dependency. All other dependencies are ignored.

When installing dependencies, the buggy-cli can work for a while and it is not clear to the user what happens. It seems stuck. Add a simple notification to indicate that a package is installing and optimally information about the progress.

Either create a server or compile to javasript and show debugging information in the interactive html page.

Information and controls that should/could be available:

• "Step through", e.g. for every (selected) atomic
• Current value of each channel
• Log all events over a channel
• Show waiting nodes
• Show different recursion levels separately
• Record / replay feature with black boxing compounds (could increase speed significantly?)

Currently lambdas are applied via apply which takes exactly one argument and one function. Partials can set one argument of a function and return a new function that takes one argument less.

It is cleaner to separate partials and calling the function. The call node represents a function call of an function without arguments, thus apply = call partial.

The CLI tasks are all implemented in cli.js. It would be handy to use them in other projects programmatically and thus there should be an API.

Creating SVGs currently leads to the usage of two different graphtools versions. This should be impossible, but it is not. I get the following toolchain:

[ { name: 'lisgy',
    module: '@buggyorg/lisgy',
    minVersion: '0.2.0-pre.1',
    consumes: 'input',
    produces: 'portgraph',
    activatedBy: [ '$<bin> input' ],
    args: 'input',
    version: '0.2.0-pre.1',
    graphtools: '0.4.0-pre.8' },
  { name: 'resolve',
    module: '@buggyorg/resolve',
    minVersion: '0.2.2',
    consumes: 'portgraph',
    produces: 'portgraph',
    activatedBy: [ 'resolve' ],
    version: '0.2.2',
    graphtools: '0.4.0-pre.11' },
  { name: 'portgraph2kgraph',
    module: '@buggyorg/portgraph2kgraph',
    consumes: 'portgraph',
    depends: [ 'resolve' ],
    produces: 'kgraph',
    version: '0.1.2',
    graphtools: '0.4.0-pre.8' },
  { name: 'graphify',
    module: '@buggyorg/graphify',
    consumes: 'kgraph',
    produces: 'svg',
    version: '0.1.27',
    graphtools: null } ]

It uses the version 0.4.0-pre.8 and 0.4.0-pre.11. There is currently no combination that can be selected. It should simply print an error indicating the impossible situation.

At the moment, buggy saves the index.html with the graph included as index2.html into the node_modules directory of graphify. This approach is not very clean and also doesn't work for those who install global npm applications with sudo.