Command line for diff for tree sequences?

[petrelharp]

It'd be nice to do something like

diff file1.trees file2.trees

and get the differences between the tables. Is there a quick (non-python) way of doing this?

[petrelharp]

Yes! tskit will do this. To get a simple example:

import msprime
ts = msprime.simulate(6, recombination_rate=1.0)
sts = ts.simplify(range(4))
ts.dump("ts.trees")
sts.dump("sts.trees")

Now, we can look at the node tables like so:

$ tskit nodes sts.trees
id	is_sample	time	population	individual	metadata
0	1	0.000000	0	-1	
1	1	0.000000	0	-1	
2	1	0.000000	0	-1	
3	1	0.000000	0	-1	
4	0	0.099054	0	-1	
5	0	0.873640	0	-1	
6	0	2.301402	0	-1	
7	0	3.290248	0	-1	
8	0	3.582909	0	-1	
9	0	3.757544	0	-1	
10	0	4.407859	0	-1

and so we can do a "diff" like this:

for x in info individuals nodes edges sites mutations populations provenances
do
    echo $x
    diff <(tskit $x ts.trees) <(tskit $x sts.trees)
done

which will tell me, for instance:

nodes
6,16c6,12
< 4	1	0.000000	0	-1	
< 5	1	0.000000	0	-1	
< 6	0	0.078407	0	-1	
< 7	0	0.099054	0	-1	
< 8	0	0.152731	0	-1	
< 9	0	0.873640	0	-1	
< 10	0	2.301402	0	-1	
< 11	0	3.290248	0	-1	
< 12	0	3.582909	0	-1	
< 13	0	3.757544	0	-1	
< 14	0	4.407859	0	-1	
---
> 4	0	0.099054	0	-1	
> 5	0	0.873640	0	-1	
> 6	0	2.301402	0	-1	
> 7	0	3.290248	0	-1	
> 8	0	3.582909	0	-1	
> 9	0	3.757544	0	-1	
> 10	0	4.407859	0	-1

(the diff here is a bit big - maybe there's a better example with a smaller diff?).

[jeromekelleher]

Nice idea! We could even add a tskit diff command using difflib under the hood...

[benjeffery]

Might be a "good first issue".

[petrelharp]

I think this is mostly useful for developers testing things - not sure when else we expect two tree sequences to be the nearly the same but not quite.