- Fix many issues that had crept in - one being (seq ds) being a sequence of keyvals which is wrong - java TMD implements seq/reduce as sequence of columns. This leads to other incompatilibities mainly keys/vals no longer work!!..
(assoc ds :b (dt/reify-reader ...))works correctly both when using :object datatype and and exact datatype - this is a lazy operation and does not rescan the data for ##NaN and friends -- if you want to provide a missing values using the form#:tech.v3.dataset{:data (reify-reader...) :missing #{missing data}}. This form works with assoc. As always passing in a container with:objectdatatype will always force a scan of the data.
- Latest TMD support - clj-transit is now in TMD and not in this project as transit is a first-class format for tmd.
- clojars coord is now com.cnuernber/tmdjs to match other libs.
- Added high performance pathway for parsing datasets and realizing intermediate datasets. The dataset-parser now implements several protocols including count, nth, so you can get, for instance, the last row during the parsing phase. When using the dataset-parser you have to use the tech.v3.dataset.protocols/-add-row and -add-rows methods.
- fast-nth fixed for integer columns.
- various fixes around parsing data.