# Problem with Interpolations. Did it change?

**URL:** <https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685>\
**Category:** New to Julia\
**Tags:** question\
**Created:** [May 6, 2021, 9:50pm UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685 "2021-05-06T21:50:00Z")\
**Posts on this page:** 15\
**Page:** 1

<div class="post-metadata">

**Author:** ![marianoarnaiz](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/marianoarnaiz/32/19377_2.png) [@marianoarnaiz](https://discourse.julialang.org/u/marianoarnaiz)\
**Post date:** [May 6, 2021, 9:50pm UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/1 "2021-05-06T21:50:01Z")

</div>

Hi everyone, I am updating a code of mine and on it I need to be able to interpolate a table. Mind that there a steps on the table (where the same value of X, has 2 values of Y1) and I used to have this lines:

```julia
    ivp = LinearInterpolation(model[:,1], model[:,2]);
    iivp = zeros(Float64, kmax); # inverse of ivp
    for k=1:kmax
        iivp[k] = 1/ivp(vz[k]);
    end

```

model is a table (see bellow) where the firsts columns had repeated values. It never gave me a problem before, but now I get:

> ERROR: knot-vectors must be unique and sorted in increasing order

Any idea?

> model=[ -8.0 5.7 3.46 2.72  
> 0.0 5.8 3.46 2.72  
> 20.0 6.0 3.46 2.72  
> 20.0 6.5 3.85 2.92  
> 35.0 6.7 3.85 2.92  
> 35.0 8.04 4.48 3.3198  
> 77.5 8.045 4.49 3.3455  
> 120.0 8.05 4.5 3.3713  
> 165.0 8.175 4.509 3.3985  
> 210.0 8.3 4.518 3.4258  
> 210.0 8.3 4.523 3.4258  
> 260.0 8.4825 4.609 3.4561  
> 310.0 8.665 4.696 3.4864  
> 360.0 8.8475 4.783 3.5167  
> 410.0 9.03 4.87 3.547  
> 410.0 9.36 5.08 3.7557  
> 460.0 9.528 5.186 3.8175  
> 510.0 9.696 5.292 3.8793  
> 560.0 9.864 5.398 3.941  
> 610.0 10.032 5.504 4.0028  
> 660.0 10.2 5.61 4.0646  
> 660.0 10.79 5.96 4.3714  
> 710.0 10.9229 6.0897 4.401  
> 760.0 11.0558 6.2095 4.4305  
> 809.5 11.1353 6.2426 4.4596  
> 859.0 11.2221 6.2798 4.4885  
> 908.5 11.3068 6.316 4.5173  
> 958.0 11.3896 6.3512 4.5459  
> 1007.5 11.4705 6.3854 4.5744  
> 1057.0 11.5495 6.4187 4.6028  
> 1106.5 11.6269 6.451 4.631  
> 1156.0 11.7026 6.4828 4.6591  
> 1205.5 11.7766 6.5138 4.687  
> 1255.0 11.8491 6.5439 4.7148  
> 1304.5 11.92 6.5727 4.7424  
> 1354.0 11.9895 6.6008 4.7699  
> 1403.5 12.0577 6.6285 4.7973  
> 1453.0 12.1245 6.6555 4.8245  
> 1502.5 12.1912 6.6815 4.8515  
> 1552.0 12.255 6.7073 4.8785  
> 1601.5 12.3185 6.7326 4.9052  
> 1651.0 12.3819 6.7573 4.9319  
> 1700.5 12.4426 6.7815 4.9584  
> 1750.0 12.5031 6.8052 4.9847  
> 1799.5 12.5631 6.8286 5.0109  
> 1849.0 12.6221 6.8515 5.037  
> 1898.5 12.6804 6.8742 5.0629  
> 1948.0 12.7382 6.8972 5.0887  
> 1997.5 12.7956 6.9194 5.1143  
> 2047.0 12.8526 6.9418 5.1398  
> 2096.5 12.9096 6.9627 5.1652  
> 2146.0 12.9668 6.9855 5.1904  
> 2195.5 13.0222 7.0063 5.2154  
> 2245.0 13.0783 7.0281 5.2403  
> 2294.5 13.1336 7.05 5.2651  
> 2344.0 13.1894 7.072 5.2898  
> 2393.5 13.2465 7.0931 5.3142  
> 2443.0 13.3018 7.1144 5.3386  
> 2492.5 13.3585 7.1369 5.3628  
> 2542.0 13.4156 7.1586 5.3869  
> 2591.5 13.4741 7.1807 5.4108  
> 2640.0 13.5312 7.2031 5.4345  
> 2690.0 13.59 7.2258 5.4582  
> 2740.0 13.6494 7.249 5.4817  
> 2740.0 13.6494 7.249 5.4817  
> 2789.67 13.653 7.2597 5.5051  
> 2839.33 13.6566 7.2704 5.5284  
> 2891.5 13.6602 7.2811 5.5515  
> 2891.5 8.0 0.0 9.9145  
> 2939.33 8.0382 0.0 9.9942  
> 2989.66 8.1283 0.0 10.0722  
> 3039.99 8.2213 0.0 10.1485  
> 3090.32 8.3122 0.0 10.2233  
> 3140.66 8.4001 0.0 10.2964  
> 3190.99 8.4861 0.0 10.3679  
> 3241.32 8.5692 0.0 10.4378  
> 3291.65 8.6496 0.0 10.5062  
> 3341.98 8.7283 0.0 10.5731  
> 3392.31 8.8036 0.0 10.6385  
> 3442.64 8.8761 0.0 10.7023  
> 3492.97 8.9461 0.0 10.7647  
> 3543.3 9.0138 0.0 10.8257  
> 3593.64 9.0792 0.0 10.8852  
> 3643.97 9.1426 0.0 10.9434  
> 3694.3 9.2042 0.0 11.0001  
> 3744.63 9.2634 0.0 11.0555  
> 3794.96 9.3205 0.0 11.1095  
> 3845.29 9.376 0.0 11.1623  
> 3895.62 9.4297 0.0 11.2137  
> 3945.95 9.4814 0.0 11.2639  
> 3996.28 9.5306 0.0 11.3127  
> 4046.62 9.5777 0.0 11.3604  
> 4096.95 9.6232 0.0 11.4069  
> 4147.28 9.6673 0.0 11.4521  
> 4197.61 9.71 0.0 11.4962  
> 4247.94 9.7513 0.0 11.5391  
> 4298.27 9.7914 0.0 11.5809  
> 4348.6 9.8304 0.0 11.6216  
> 4398.93 9.8682 0.0 11.6612  
> 4449.26 9.9051 0.0 11.6998  
> 4499.6 9.941 0.0 11.7373  
> 4549.93 9.9761 0.0 11.7737  
> 4600.26 10.0103 0.0 11.8092  
> 4650.59 10.0439 0.0 11.8437  
> 4700.92 10.0768 0.0 11.8772  
> 4801.58 10.1415 0.0 11.9414  
> 4851.91 10.1739 0.0 11.9722  
> 4902.24 10.2049 0.0 12.0001  
> 4952.58 10.2329 0.0 12.0311  
> 5002.91 10.2565 0.0 12.0593  
> 5053.24 10.2745 0.0 12.0867  
> 5103.57 10.2854 0.0 12.1133  
> 5153.5 10.289 0.0 12.1391  
> 5153.5 11.0427 3.5043 12.7037  
> 5204.61 11.0585 3.5187 12.7289  
> 5255.32 11.0718 3.5314 12.753  
> 5306.04 11.085 3.5435 12.776  
> 5356.75 11.0983 3.5551 12.798  
> 5407.46 11.1166 3.5661 12.8188  
> 5458.17 11.1316 3.5765 12.8387  
> 5508.89 11.1457 3.5864 12.8574  
> 5559.6 11.159 3.5957 12.8751  
> 5610.31 11.1715 3.6044 12.8917  
> 5661.02 11.1832 3.6126 12.9072  
> 5711.74 11.1941 3.6202 12.9217  
> 5813.16 11.2134 3.6337 12.9474  
> 5863.87 11.2219 3.6396 12.9586  
> 5914.59 11.2295 3.645 12.9688  
> 5965.3 11.2364 3.6498 12.9779  
> 6016.01 11.2424 3.654 12.9859  
> 6066.72 11.2477 3.6577 12.9929  
> 6117.44 11.2521 3.6608 12.9988  
> 6168.15 11.2557 3.6633 13.0036  
> 6218.86 11.2586 3.6653 13.0074  
> 6269.57 11.2606 3.6667 13.01  
> 6320.29 11.2618 3.6675 13.0117  
> 6371.0 11.2622 3.6678 13.0122]

---

<div class="post-metadata">

**Author:** ![stillyslalom](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/stillyslalom/32/45687_2.png) [@stillyslalom](https://discourse.julialang.org/u/stillyslalom)\
**Post date:** [May 6, 2021, 10:31pm UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/2 "2021-05-06T22:31:27Z")

</div>

As the error says, your knot vector can’t have repeated values. It’s ill-determined: do you pick one of the values at random? Take the mean? You’ll need to decide how you want the interpolation to behave, and clean your data appropriately to remove repeated values.

This change was introduced in [this commit](https://github.com/JuliaMath/Interpolations.jl/commit/318ebc88ca1fc084754f3c741266537f901a3310), so in the short term, you can just pin the version to the prior release (v0.13.1).

---

<div class="post-metadata">

**Author:** ![marianoarnaiz](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/marianoarnaiz/32/19377_2.png) [@marianoarnaiz](https://discourse.julialang.org/u/marianoarnaiz)\
**Post date:** [May 7, 2021, 8:56am UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/3 "2021-05-07T08:56:54Z")

</div>

This is very inconvinient … most of the data I work with looks like this.  
Is there a way for me to “grab” a copy of the old interpolation so I can keep it and not be bound to the Interpolations package?

Or maybe it could become an option in the package.

---

<div class="post-metadata">

**Author:** ![FPGro](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/fpgro/32/20822_2.png) [@FPGro](https://discourse.julialang.org/u/FPGro)\
**Post date:** [May 7, 2021, 11:43am UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/4 "2021-05-07T11:43:26Z")

</div>

The package is MIT expat licensed, so you have

> […] the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:  
> The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. […]

So yeah, if you feel the need to copy the state of Interpolations pre v0.13.1 you can do so. You can even fork it I believe and install from your own repository. Just don’t try to register it in General 🙂

---

<div class="post-metadata">

**Author:** ![pfitzseb](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/pfitzseb/32/45566_2.png) [@pfitzseb](https://discourse.julialang.org/u/pfitzseb)\
**Post date:** [May 7, 2021, 12:00pm UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/5 "2021-05-07T12:00:40Z")

</div>

I’d very much recommend properly preprocessing your data though.

---

<div class="post-metadata">

**Author:** ![marianoarnaiz](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/marianoarnaiz/32/19377_2.png) [@marianoarnaiz](https://discourse.julialang.org/u/marianoarnaiz)\
**Post date:** [May 7, 2021, 12:09pm UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/6 "2021-05-07T12:09:21Z")

</div>

Given that it is a conceptual model of Earth’s structure, I rather not change it. But I welcome any suggestions

---

<div class="post-metadata">

**Author:** ![marianoarnaiz](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/marianoarnaiz/32/19377_2.png) [@marianoarnaiz](https://discourse.julialang.org/u/marianoarnaiz)\
**Post date:** [May 7, 2021, 12:09pm UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/7 "2021-05-07T12:09:56Z")

</div>

Thanks. Maybe I will find how to fork it

---

<div class="post-metadata">

**Author:** ![nilshg](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/nilshg/32/2283_2.png) [@nilshg](https://discourse.julialang.org/u/nilshg)\
**Post date:** [May 7, 2021, 12:54pm UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/8 "2021-05-07T12:54:07Z")

</div>

If you want to use version 0.13.1 you can just pin it at that version, no need to fork?

In terms of changing the model, I guess the first step would be to see what Interpolations 0.13 does when there are duplicates in the knot vector, and whether this is actually what you expect (the question above is quite relevant I think, it’s not clearly to me what _should_ be happening when you present Interpolations with such data).

---

<div class="post-metadata">

**Author:** ![pfitzseb](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/pfitzseb/32/45566_2.png) [@pfitzseb](https://discourse.julialang.org/u/pfitzseb)\
**Post date:** [May 7, 2021, 1:00pm UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/9 "2021-05-07T13:00:13Z")

</div>

But Interpolations already made _a_ choice for you in past versions (afaict it just takes the first value). So now you’ll just need to do the same manually, with the bonus that you actually know what’s going on.

---

<div class="post-metadata">

**Author:** ![marianoarnaiz](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/marianoarnaiz/32/19377_2.png) [@marianoarnaiz](https://discourse.julialang.org/u/marianoarnaiz)\
**Post date:** [May 7, 2021, 1:15pm UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/10 "2021-05-07T13:15:11Z")

</div>

Here is a quick example @nilshg

It does choose the first value for the knot as @pfitzseb says, but understand that right after is another:

```julia

julia> ivp(19.999)
5.9999899999999995

julia> ivp(20.001)
6.5000133333333325

julia> ivp(20)
6.0

```

This is exactly what I need, but the option of choosing the value from the second line or the average (as a value of the knot) would be cool to have.

---

<div class="post-metadata">

**Author:** ![nilshg](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/nilshg/32/2283_2.png) [@nilshg](https://discourse.julialang.org/u/nilshg)\
**Post date:** [May 7, 2021, 1:48pm UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/11 "2021-05-07T13:48:54Z")

</div>

That does sounds like something that should be happening in pre-processing: basically you stick your data in a DataFrame and then do something like

```julia
combine(groupby(data, :knots, sort = true), observations => first)

```

Then you pass the result of this to `itp`. Of course you can then replace `first` with `mean`, `last`, or whatever other function you might need.

---

<div class="post-metadata">

**Author:** ![stillyslalom](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/stillyslalom/32/45687_2.png) [@stillyslalom](https://discourse.julialang.org/u/stillyslalom)\
**Post date:** [May 8, 2021, 4:41am UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/12 "2021-05-08T04:41:42Z")

</div>

In fairness, the underlying PR is (arguably) a breaking change, and should have coincided with a version bump, but it wasn’t even mentioned in any of the release notes (afaict).The old behavior was ambiguous, but it wasn’t incorrect, per se, and this change should have been more clearly publicized.

---

<div class="post-metadata">

**Author:** ![marianoarnaiz](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/marianoarnaiz/32/19377_2.png) [@marianoarnaiz](https://discourse.julialang.org/u/marianoarnaiz)\
**Post date:** [May 8, 2021, 9:34am UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/13 "2021-05-08T09:34:58Z")

</div>

I will have to give this a try. Is data here a vector?

---

<div class="post-metadata">

**Author:** ![marianoarnaiz](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/marianoarnaiz/32/19377_2.png) [@marianoarnaiz](https://discourse.julialang.org/u/marianoarnaiz)\
**Post date:** [May 8, 2021, 9:36am UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/14 "2021-05-08T09:36:02Z")

</div>

Definitely.  
Also, in the spirit of making the package as flexible as possible I think keeping it as a option is a good idea. Maybe a little function can be added to preprocess the data for this kind of case.

---

<div class="post-metadata">

**Author:** ![nilshg](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/nilshg/32/2283_2.png) [@nilshg](https://discourse.julialang.org/u/nilshg)\
**Post date:** [May 8, 2021, 11:35am UTC](https://discourse.julialang.org/t/problem-with-interpolations-did-it-change/60685/15 "2021-05-08T11:35:03Z")

</div>

In my example `data` is a DataFrame - I’m not sure what your actual data structure looks like, but I was thinking of something like this:

```julia
julia> using DataFrames, Statistics

julia> data = DataFrame(knots = [1, 2, 2, 3], obs1 = rand(4), obs2 = rand(4))
4×3 DataFrame
 Row │ knots obs1 obs2     
     │ Int64 Float64 Float64  
─────┼────────────────────────────
   1 │ 1 0.817886 0.915149
   2 │ 2 0.604035 0.323521
   3 │ 2 0.242594 0.243895
   4 │ 3 0.0296911 0.494106

julia> observations = [:obs1, :obs2];

julia> combine(groupby(data, :knots), observations .=> first .=> observations)
3×3 DataFrame
 Row │ knots obs1 obs2     
     │ Int64 Float64 Float64  
─────┼────────────────────────────
   1 │ 1 0.817886 0.915149
   2 │ 2 0.604035 0.323521
   3 │ 3 0.0296911 0.494106

julia> combine(groupby(data, :knots), observations .=> mean .=> observations)
3×3 DataFrame
 Row │ knots obs1 obs2     
     │ Int64 Float64 Float64  
─────┼────────────────────────────
   1 │ 1 0.817886 0.915149
   2 │ 2 0.423314 0.283708
   3 │ 3 0.0296911 0.494106

```

(in your actual use case `knots` might be more than one column if you’re sampling on a higher dimensional grid, but you hopefully get the gist)