Documentation request: Row and column selection and iteration examples about anndata HOT 3 CLOSED

Of course, asking for help is often the fastest way to find the answer on your own. This works:

selected = adata[:, adata.var_names.isin(['Tcea1',])]

# This is for a complete matrix
for g, gene in enumerate(selected.var_names):
    for c, cell in enumerate(selected.obs_names):
        print("{0}\t{1}\t{2}".format(cell, gene, selected.X[c]))

It seems like using "isin()" when you're only going after one gene is suboptimal though.

from anndata.

Hi Joshua!
When you select one row or column, the matrix will get flattened, as this is numpy's behavior and often very practical. See the following comment in the Notes section http://anndata.readthedocs.io/en/latest/anndata.AnnData.html

adata = AnnData(np.ones((2, 2)))
adata[:, 0].X == adata.X[:, 0]

So, if you're only going for single gene, then you get a 1d vector and not a matrix. Would you like to see this behavior different? Your trick with ['Tcea1', ] is definitely a very good workaround if you want something two-dimensional.

PS: You can also use 'Tcea1' == adata.var_names instead of using .isin(...) when you only have a single element.

from anndata.

Of course, asking for help is often the fastest way to find the answer on your own.

haha yeah, it’s called rubber duck debugging

When you select one row or column, the matrix will get flattened, as this is numpy's behavior and often very practical.

Oh, once upon a time I made sure that this doesn’t happen in AnnData, since AnnData is always a 2D structure (You should always have (at least one) named cell(s) and gene(s), never just a 1D vector of expression values without knowing what dimension it is along)

from anndata.