Recompilation (and delay) on every new run seems like a major issue. Probably, if you call `Mix.install/?`, then you should pass the target directory to cache compiled dependencies.
By default it can use `_build` as all current mix projects. > On 21 Jan 2021, at 17:48, José Valim <jose.va...@dashbit.co> wrote: > > > The latter solution could most easily be solved by using /tmp and making a > > new folder each iex instance that calls mix install (possibly also > > respecting the cache dir flag options) > > Good points. > > That's kind of how the current implementation works. The cache is in tmp and > the cache key is the MD5 of the deps given to install. This means that the > cache is most likely to be used only by the current script. So it reduces the > odds of interference between scripts. Otherwise you will have to fetch and > compile dependencies every time the script runs, which would be too slow. > > Issues like Elixir version and Erlang version can be solved in two ways: 1. > don't do anything because Mix already solves those or 2. add those to the > cache key - either as part of the directory structure (my vote) or as part of > the MD5 computation. > > We could also support explicit naming for better control over the cache - but > I would wait for those until people effectively run into cache issues, if > ever. :) > > > > On Thu, Jan 21, 2021 at 3:35 PM Jonathan Arnett <jonarnet...@gmail.com > <mailto:jonarnet...@gmail.com>> wrote: > I am also a fan of the proposal. One addendum that I would like to suggest is > adding a flag to IEx such that a library can be "installed" when the shell > starts. > > e.g. > iex --install :decimal > > On 1/21/21 9:01 AM, Anil Kulkarni wrote: >> I love the general idea. I think it's a great solution for copy/paste/run >> needs that a scripting language should have. >> >> I do have a question about the caching packages across runs. What is the >> motivation? Is it to increase performance, or are there other reasons? >> >> My concern is that inevitably there will be issues with the cache. It >> reminds me of global pip requirements, global npm installs etc. >> >> Some issues I have personally run into with these solutions ( python and js) >> include: >> >> 1) the cache size increasing and needing to manually trim it. Consider >> popping a shell into a production environment to fix an issue. You >> wouldn'twant to leave stuff behind. >> >> 2) Managing dependencies across language upgrades (e.g. updating node to >> 15.5, what happens to the dependencies? What happens for sxs (side by side) >> >> 3. Inconsistencies due to caching, especially with node. E.g. Needing to >> blow away the cache and download fresh to fix an issue >> >> 4. Fire corruption due to other programs needing to nuke the cache >> >> 5. Issues that arise when backing up and then trying to restore to a >> different home folder (I had a python dependency hard code the path an >> install) >> >> 6. File permission changes, especially when switching between different >> users on chroot'd terminals >> >> Some of these issues are going to be a problem anyways, but I think that >> caching provides another avenue for things go wrong. >> >> My suggestion would be to either add explicit cache management functionality >> to elixir - including which path it lives, and blowing away the cache or to >> not use one for now. >> >> The former could look like this: >> ELIXIR_CACHE_DIR=/foo iex >> (eg set an environment var then call iex) or with parameters passed in iex >> --cache-dir >> iex --clear-cache >> >> The latter solution could most easily be solved by using /tmp and making a >> new folder each iex instance that calls mix install (possibly also >> respecting the cache dir flag options) >> >> -Anil >> >>> On Jan 21, 2021, at 05:02, Felipe Stival <v0id...@gmail.com> >>> <mailto:v0id...@gmail.com> wrote: >>> >>> >>> +1 for this proposal. >>> >>> On Thu, Jan 21, 2021, 09:54 Jon Rowe <m...@jonrowe.co.uk >>> <mailto:m...@jonrowe.co.uk>> wrote: >>> I plus one this proposal, it’s similar (for scripts at least) to what Ruby >>> has with bundler inline mode, I like this because such techniques allow you >>> to create single file reproductions of issues, which are great for bug >>> reports in PRs. >>> >>> Cheers >>> Jon >>> ---------------- >>> m...@jonrowe.co.uk <mailto:m...@jonrowe.co.uk> >>> https://jonrowe.co.uk <https://jonrowe.co.uk/> >>> >>> On 21 January 2021 at 12:18, eksperimental wrote: >>>> `Mix.install(:decimal)` should also work out of the box IMO. >>>> >>>> On Thu, 21 Jan 2021 03:18:44 -0800 (PST) >>>> "woj...@wojtekmach.pl <mailto:woj...@wojtekmach.pl>" <woj...@wojtekmach.pl >>>> <mailto:woj...@wojtekmach.pl>> wrote: >>> -- >>> You received this message because you are subscribed to the Google Groups >>> "elixir-lang-core" group. >>> To unsubscribe from this group and stop receiving emails from it, send an >>> email to elixir-lang-core+unsubscr...@googlegroups.com >>> <mailto:elixir-lang-core+unsubscr...@googlegroups.com>. >>> To view this discussion on the web visit >>> https://groups.google.com/d/msgid/elixir-lang-core/dejalu-217-a7302266-156c-46d8-86f9-089e2065af2b%40jonrowe.co.uk >>> >>> <https://groups.google.com/d/msgid/elixir-lang-core/dejalu-217-a7302266-156c-46d8-86f9-089e2065af2b%40jonrowe.co.uk?utm_medium=email&utm_source=footer>. >>> -- >>> You received this message because you are subscribed to the Google Groups >>> "elixir-lang-core" group. >>> To unsubscribe from this group and stop receiving emails from it, send an >>> email to elixir-lang-core+unsubscr...@googlegroups.com >>> <mailto:elixir-lang-core+unsubscr...@googlegroups.com>. >>> To view this discussion on the web visit >>> https://groups.google.com/d/msgid/elixir-lang-core/CAKC64%2ByRtG-XGWkSADPXHzOLZ8AJ4g73DxEjC_8Jp1b3-sHKAg%40mail.gmail.com >>> >>> <https://groups.google.com/d/msgid/elixir-lang-core/CAKC64%2ByRtG-XGWkSADPXHzOLZ8AJ4g73DxEjC_8Jp1b3-sHKAg%40mail.gmail.com?utm_medium=email&utm_source=footer>. >> -- >> You received this message because you are subscribed to the Google Groups >> "elixir-lang-core" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to elixir-lang-core+unsubscr...@googlegroups.com >> <mailto:elixir-lang-core+unsubscr...@googlegroups.com>. >> To view this discussion on the web visit >> https://groups.google.com/d/msgid/elixir-lang-core/0100017725401617-8cdc5c5f-8434-42d6-a3b5-0272cbf65824-000000%40email.amazonses.com >> >> <https://groups.google.com/d/msgid/elixir-lang-core/0100017725401617-8cdc5c5f-8434-42d6-a3b5-0272cbf65824-000000%40email.amazonses.com?utm_medium=email&utm_source=footer>. > > > -- > You received this message because you are subscribed to the Google Groups > "elixir-lang-core" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to elixir-lang-core+unsubscr...@googlegroups.com > <mailto:elixir-lang-core+unsubscr...@googlegroups.com>. > To view this discussion on the web visit > https://groups.google.com/d/msgid/elixir-lang-core/62020d79-4f44-38d5-65e7-a786808da633%40gmail.com > > <https://groups.google.com/d/msgid/elixir-lang-core/62020d79-4f44-38d5-65e7-a786808da633%40gmail.com?utm_medium=email&utm_source=footer>. > > -- > You received this message because you are subscribed to the Google Groups > "elixir-lang-core" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to elixir-lang-core+unsubscr...@googlegroups.com > <mailto:elixir-lang-core+unsubscr...@googlegroups.com>. > To view this discussion on the web visit > https://groups.google.com/d/msgid/elixir-lang-core/CAGnRm4JFHs6stz5sHvFabZXP%2BS3HXMLN8PKd%2BVNJ%3DmwbF8-J_A%40mail.gmail.com > > <https://groups.google.com/d/msgid/elixir-lang-core/CAGnRm4JFHs6stz5sHvFabZXP%2BS3HXMLN8PKd%2BVNJ%3DmwbF8-J_A%40mail.gmail.com?utm_medium=email&utm_source=footer>. -- You received this message because you are subscribed to the Google Groups "elixir-lang-core" group. To unsubscribe from this group and stop receiving emails from it, send an email to elixir-lang-core+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/elixir-lang-core/1A4436D3-17AC-41BB-94CB-57F45D0ECF37%40achempion.com.
