blahah / biopsy Goto Github PK

Add parsing of flags from the yaml file so that an option can be optimised to be either on or off.

target_dirs should contain:

definitions
constructors

some domains may not want to implement it, so we should only enforce it if it has been defined

assemblotron test failure:

ERROR (0:00:00.003) test: SoapDenovoTrans constructor should automatically include defaults. 
 undefined method `include?' for nil:NilClass
 @ /Users/rs404/.rvm/gems/ruby-1.9.3-p362/gems/biopsy-0.1.0.alpha/lib/biopsy/domain.rb:88:in `block in validate_target_filetypes'

caused by domain definition loading as:

{"min"=>2, "allowed_extensions"=>[".fastq", ".fq", ".fasta", ".fa"]}

keys are expected to be symbols

For some parameter spaces, convergence can take a long time depending on stochastic elements of the exploration. We should allow users to specify a time limit and return the best solution reached at that time.

It should be possible to exactly repeat a run

let's get it up to an A on code climate

Currently, objectives are run for the output of every target run. In some use cases (e.g. transcriptome assembly), multiple different parameter sets will result in exactly the same assembly. Let's take md5 hashes of all output files, and use a set of those hashes as the key in a hash of objective scores. Could potentially shave hours off a run for computationally intensive objectives.

currently target definitions must include an array of possible values for each parameter. We should allow specification of a hash for each parameter, with min, max and step of a range.

Currently domain validation errors always refer to "input files", even when validating output.