install_requires.txt

numpy>=1.8
scipy>=0.13

# Matplotlib is used in several places
# - The testing plots in lib/plots/p_p_plot.py.
# - plotf is matplotlib-backed via ggplot.
# - Several examples in the examples/ tree (Brownian motion,
#   crp_2d_demo, dpmm) use matplotlib directly.

matplotlib>=1.3.1

# We would depend on ggplot, because of the plotf facility, but ggplot
# is dead, so we forked it into the external/ directory.

# Patsy is only needed for ggplot, which backs plotf

patsy>=0.2 # Transitive through included ggplot

# Pandas is used in a couple of places
# - The profiler produces results as a DataFrame (Ripl.profile_data
#   method), though could presumably be taught not to do that.
# - plotf uses pandas as a data interchange format, because ggplot
#   does.
# - The dataset accumulation facility (`accumulate_dataset` and co)
#   has a direct dependency on pandas, in the form of the asPandas
#   method.  Don't know why it might be useful to have dataset
#   accumulation without plotf, but if so, could get rid of the
#   dependency by dropping the asPandas method.
# - The pipits example also uses pandas directly.

pandas>=0.14, <0.19 # 0.19 requires cython to build.  Don't know why pip decides to build pandas.

# Debug pictures of scaffolds
networkx

# Ripl server
Flask>=0.10
requests>=1.2

# Ubuntu 14.04 apparently needs this mentioned for notebooks to work
jsonschema

# We also have private forks of plex and lemonade, which are the
# scanner and parser generator for the surface syntax.