Message ID | 20211119043647.1251416-1-art@khadas.com |
---|---|
Headers | show |
Series | env: setenv add resolve value option | expand |
Dear Artem, In message <20211119043647.1251416-1-art@khadas.com> you wrote: > Add possibility setup env variable with additional resolving vars inside > value. Hm... if you want to evaluate variables, you should not prevent the shell to do that by enclosing them in apostrophes? > Usage examples: > > => setenv a hello; setenv b world; setenv c '${a} ${b}' > => setenv -r d '${c}! ${a}...' > => printenv d > d=hello world! hello... Without any new code added: => setenv a hello; setenv b world; setenv c ${a} ${b} => setenv d ${c}! ${a}... => printenv d d=hello world! hello... I know very well that this does not cover all use cases you might have in mind, as you speak of "deep resolve". But then, I'm first missing an explanation (and documentation) of what "deep resolve" actually means in this context, i. e. how many levels down you evaluat. Oh... the code has "int max_loop = 32;" - this is a limitation mentioned nowhere. And are you really sure this is a clever idea? I am not convinced. If we do something like this, we should not invent a new non-standard way. We should implement standard shell behaviour instead (i. e. for example something like the eval command). > Artem Lapkin (2): > env: setenv add resolve value option > test: env: deep resolve value testing This is one more of the patches that actually try to fix existing problems with our ancient verion of the hush shell. As far as I understand, Francis (added to Cc:) has started working on an update of hush to a current version. We should rather help him with that instead of implementing non-standard workarounds. Best regards, Wolfgang Denk
hi Wolfgang tnx for your comments On Fri, Nov 19, 2021 at 3:49 PM Wolfgang Denk <wd@denx.de> wrote: > > Dear Artem, > > In message <20211119043647.1251416-1-art@khadas.com> you wrote: > > Add possibility setup env variable with additional resolving vars inside > > value. > > Hm... if you want to evaluate variables, you should not prevent the > shell to do that by enclosing them in apostrophes? > next examples just demonstrate how its works for already defined env variables which contain other variables (like storred env variables) > > Usage examples: > > > > => setenv a hello; setenv b world; setenv c '${a} ${b}' > > => setenv -r d '${c}! ${a}...' > > => printenv d > > d=hello world! hello... > > Without any new code added: > sure I know about this ! see my prev message please . > => setenv a hello; setenv b world; setenv c ${a} ${b} > => setenv d ${c}! ${a}... > => printenv d > d=hello world! hello... > > > I know very well that this does not cover all use cases you might Why not have this new opportunity ? > have in mind, as you speak of "deep resolve". But then, I'm first > missing an explanation (and documentation) of what "deep resolve" recurrent resolving for variables > actually means in this context, i. e. how many levels down you > evaluat. Oh... the code has "int max_loop = 32;" - this is a i think its will be enough > limitation mentioned nowhere. And are you really sure this is a > clever idea? I am not convinced. If we do something like this, we > should not invent a new non-standard way. We should implement I believe it's a good idea ;-) > standard shell behaviour instead (i. e. for example something like > the eval command). 1) this option did not broke any exist compatibilities 2) there we talk not only about uboot shell, same time will be useful to have env_resolve for internal c usage, because env_set dont have this feature > > > Artem Lapkin (2): > > env: setenv add resolve value option > > test: env: deep resolve value testing > > This is one more of the patches that actually try to fix existing > problems with our ancient verion of the hush shell. As far as I > understand, Francis (added to Cc:) has started working on an update > of hush to a current version. We should rather help him with that > instead of implementing non-standard workarounds. yes i'm informed about this plans (and think its happens not so soon - but i provide some simple elegant solution already) but again we dont have env_resolve for internal c usage which must be very useful for example: pxe_utils.c: and some other places env_set("bootargs", finalbootargs); can be replaced by env_resolve("bootargs", ..... ) will be easy get useful features via simple solution ( deep resolve all vars by one line ) > > Best regards, > Big tnx again ! and waiting any other ideas about this > Wolfgang Denk > > -- > DENX Software Engineering GmbH, Managing Director: Wolfgang Denk > HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany > Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de > When choosing between two evils, I always like to take the one I've > never tried before. -- Mae West, "Klondike Annie"
Dear Artem, In message <CAKaHn9KZg13r=pGCo=Lv69pBP-s-EW4uxjBVuH9veT+o5j9v+g@mail.gmail.com> you wrote: > > next examples just demonstrate how its works for already defined env > variables which contain other variables (like storred env variables) Which next examples? > sure I know about this ! see my prev message please . Which exact message are you referring to here? > Why not have this new opportunity ? I think the suggested code is adding more problems than it solves. > > have in mind, as you speak of "deep resolve". But then, I'm first > > missing an explanation (and documentation) of what "deep resolve" > > recurrent resolving for variables Your implementation of recursion has an arbiotrary and undocumented depth limit. Also, I cannot see a way to prevent resolving in case I want to keep something like "$foo" in the result. But that's to be expected from such a non-standard way. Why don't you stick with what "eval" in a standard shell does? > > actually means in this context, i. e. how many levels down you > > evaluat. Oh... the code has "int max_loop = 32;" - this is a > > i think its will be enough It is a reallybad habt to implement code with arbitrary limits, as it will blow into your face (or more likely that of an innocent user) rather sooner than later. It's even worse that this limit is nowhere documented. > 1) this option did not broke any exist compatibilities > 2) there we talk not only about uboot shell, same time will be useful > to have env_resolve for internal c usage, because env_set dont have > this feature I did not say that an "eval" like construct would not be useful. But uncontrolled recursion with an undocumented depth limit is a problem. > yes i'm informed about this plans (and think its happens not so > soon - but i provide some simple elegant solution already) > but again we dont have env_resolve for internal c usage which must be > very useful On the CLI, we use the "run" command to get the desired effect. Yes, this is neither perfect nor elegant. But you can use that in C code as well. > will be easy get useful features via simple solution ( deep resolve > all vars by one line ) I understand what you want, but this is not a good way to solve the problem. I'd really rather see such efforts invested in helping Francis with the hush update - which will make such code unnecessary. Best regards, Wolfgang Denk
On Sat, Nov 20, 2021 at 8:36 PM Wolfgang Denk <wd@denx.de> wrote: > > Dear Artem, > > In message <CAKaHn9KZg13r=pGCo=Lv69pBP-s-EW4uxjBVuH9veT+o5j9v+g@mail.gmail.com> you wrote: > > > > next examples just demonstrate how its works for already defined env > > variables which contain other variables (like storred env variables) > > Which next examples? Usage examples (from commit message): => setenv a hello; setenv b world; setenv c '${a} ${b}' => setenv -r d '${c}! ${a}...' => printenv d d=hello world! hello... > > > sure I know about this ! see my prev message please . > > Which exact message are you referring to here? > > > Why not have this new opportunity ? > > I think the suggested code is adding more problems than it solves. > > > > have in mind, as you speak of "deep resolve". But then, I'm first > > > missing an explanation (and documentation) of what "deep resolve" > > > > recurrent resolving for variables > > Your implementation of recursion has an arbiotrary and undocumented > depth limit. Also, I cannot see a way to prevent resolving in case I > want to keep something like "$foo" in the result. > > But that's to be expected from such a non-standard way. > > Why don't you stick with what "eval" in a standard shell does? > > > > actually means in this context, i. e. how many levels down you > > > evaluat. Oh... the code has "int max_loop = 32;" - this is a > > > > i think its will be enough > > It is a reallybad habt to implement code with arbitrary limits, as > it will blow into your face (or more likely that of an innocent > user) rather sooner than later. It's even worse that this limit is > nowhere documented. > > > 1) this option did not broke any exist compatibilities > > 2) there we talk not only about uboot shell, same time will be useful > > to have env_resolve for internal c usage, because env_set dont have > > this feature > > I did not say that an "eval" like construct would not be useful. > But uncontrolled recursion with an undocumented depth limit is > a problem. > > > yes i'm informed about this plans (and think its happens not so > > soon - but i provide some simple elegant solution already) > > but again we dont have env_resolve for internal c usage which must be > > very useful > > On the CLI, we use the "run" command to get the desired effect. Yes, > this is neither perfect nor elegant. But you can use that in C code > as well. > > > will be easy get useful features via simple solution ( deep resolve > > all vars by one line ) > > I understand what you want, but this is not a good way to solve the > problem. I'd really rather see such efforts invested in helping > Francis with the hush update - which will make such code unnecessary. > > Best regards, > > Wolfgang Denk > > -- > DENX Software Engineering GmbH, Managing Director: Wolfgang Denk > HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany > Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de > Success in marriage is not so much finding the right person as it is > being the right person.
On Sat, Nov 20, 2021 at 8:36 PM Wolfgang Denk <wd@denx.de> wrote: > > Dear Artem, > > In message <CAKaHn9KZg13r=pGCo=Lv69pBP-s-EW4uxjBVuH9veT+o5j9v+g@mail.gmail.com> you wrote: > > > > next examples just demonstrate how its works for already defined env > > variables which contain other variables (like storred env variables) > > Which next examples? > > > sure I know about this ! see my prev message please . > > Which exact message are you referring to here? > > > Why not have this new opportunity ? > > I think the suggested code is adding more problems than it solves. > > > > have in mind, as you speak of "deep resolve". But then, I'm first > > > missing an explanation (and documentation) of what "deep resolve" > > > > recurrent resolving for variables > > Your implementation of recursion has an arbiotrary and undocumented > depth limit. we can improve it if it will be really necessary > Also, I cannot see a way to prevent resolving in case I > want to keep something like "$foo" in the result. > `setenv -r` resolve only bracked vars like ${VAR} if u need keen some $VAR just use it without brackets or just use same setenv without -r option - whats a problem ? > But that's to be expected from such a non-standard way. > we can use setenv with -r or without this option (default standard way) again didn't see problems (setenv -r is special option especially for non standard way ) > Why don't you stick with what "eval" in a standard shell does? > show me please how can i do it via standard way maybe i miss something ? > > > actually means in this context, i. e. how many levels down you > > > evaluat. Oh... the code has "int max_loop = 32;" - this is a > > > > i think its will be enough > > It is a reallybad habt to implement code with arbitrary limits, as > it will blow into your face (or more likely that of an innocent > user) rather sooner than later. It's even worse that this limit is > nowhere documented. > > > 1) this option did not broke any exist compatibilities > > 2) there we talk not only about uboot shell, same time will be useful > > to have env_resolve for internal c usage, because env_set dont have > > this feature > > I did not say that an "eval" like construct would not be useful. > But uncontrolled recursion with an undocumented depth limit is > a problem. > > > yes i'm informed about this plans (and think its happens not so > > soon - but i provide some simple elegant solution already) > > but again we dont have env_resolve for internal c usage which must be > > very useful > > On the CLI, we use the "run" command to get the desired effect. Yes, > this is neither perfect nor elegant. But you can use that in C code > as well. > > > will be easy get useful features via simple solution ( deep resolve > > all vars by one line ) > > I understand what you want, but this is not a good way to solve the > problem. I'd really rather see such efforts invested in helping > Francis with the hush update - which will make such code unnecessary. > > Best regards, > > Wolfgang Denk > > -- > DENX Software Engineering GmbH, Managing Director: Wolfgang Denk > HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany > Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de > Success in marriage is not so much finding the right person as it is > being the right person.
Dear Artem, In message <CAKaHn9+yL3naiDqbsQJWQ0Po=g=TmjjcPSKxkx5C4ay1M1AsUA@mail.gmail.com> you wrote: > > > > > > next examples just demonstrate how its works for already defined env > > > variables which contain other variables (like storred env variables) > > > > Which next examples? > > Usage examples (from commit message): > > => setenv a hello; setenv b world; setenv c '${a} ${b}' > => setenv -r d '${c}! ${a}...' > => printenv d > d=hello world! hello... This is a very simple example, and I showed you how you can solve this one by just omitting the apostrophes: => setenv a hello; setenv b world; setenv c ${a} ${b} => setenv d ${c}! ${a}... => printenv d d=hello world! hello... I _think_ what you actually have in mind is something like this: => setenv a hello => setenv b world => setenv c '${a} ${b}' => setenv a goodbye => setenv b sunshine something to set d to: '${c}! ${a}...' => printenv d Here my simple approach does not show what you want to have: => setenv a hello => setenv b world => setenv c ${a} ${b} => setenv a goodbye => setenv b sunshine => setenv d ${c}! ${a}... => printenv d d=hello world! goodbye... That's because here evaluation takes place at assignment, but you want it when used - but being recursive here is neither a good idea nor standard. How would you do it in a standard posix shell? You would have to use "eval", like that: $ a=hello $ b=world $ c='${a} ${b}' $ a=goodbye $ b=sunshine $ d=$(eval echo $c) $ echo $d goodbye sunshine But please note that "eval" is _not_ recursive!! $ a='$b' $ eval echo $c $b sunshine And this is why I object against this patch. Oh, and in U-Boot you could write this as: => setenv a hello => setenv b world => setenv c '${a} ${b}' => setenv a goodbye => setenv b sunshine => setenv foo "setenv d ${c}! ${a}..." => run foo => printenv d d=goodbye sunshine! goodbye... And yes, here you have to be careful about using ' or " as there is no recursion like you might expect. So yes, it would be nice if we had "eval" (which will ocme with the hush update), and no, "eval" does not recurse either. Best regards, Wolfgang Denk
On Mon, Nov 22, 2021 at 4:52 PM Wolfgang Denk <wd@denx.de> wrote: > > Dear Artem, > > In message <CAKaHn9+yL3naiDqbsQJWQ0Po=g=TmjjcPSKxkx5C4ay1M1AsUA@mail.gmail.com> you wrote: > > > > > > > > next examples just demonstrate how its works for already defined env > > > > variables which contain other variables (like storred env variables) > > > > > > Which next examples? > > > > Usage examples (from commit message): > > > > => setenv a hello; setenv b world; setenv c '${a} ${b}' > > => setenv -r d '${c}! ${a}...' > > => printenv d > > d=hello world! hello... > > This is a very simple example, and I showed you how you can solve > this one by just omitting the apostrophes: > > => setenv a hello; setenv b world; setenv c ${a} ${b} > => setenv d ${c}! ${a}... > => printenv d > d=hello world! hello... > sure i know it and its easy, but we still have some misunderstanding, and i will try explain again our problem our problem (TASK) we have: some env variables which already defined and we totally dont know about content ( for example its was improved by `env import` or loaded from env storage and "setenv a hello; setenv b world; setenv c '${a} ${b}'; setenv d '${c}! ${a}...' ;" just example how emulate this env state ) we need: setup/resolve env variable for example from d (and need resolve all sub included vars) => setenv -r e $d => printenv e => e=hello world! hello... please explain how we can get it without -r (deep resolve option) via standard way ? > > I _think_ what you actually have in mind is something like this: > > => setenv a hello > => setenv b world > => setenv c '${a} ${b}' > => setenv a goodbye > => setenv b sunshine > > something to set d to: '${c}! ${a}...' > > => printenv d > > Here my simple approach does not show what you want to have: > > => setenv a hello > => setenv b world > => setenv c ${a} ${b} > => setenv a goodbye > => setenv b sunshine > => setenv d ${c}! ${a}... > => printenv d > d=hello world! goodbye... > > That's because here evaluation takes place at assignment, but you > want it when used - but being recursive here is neither a good idea > nor standard. > > How would you do it in a standard posix shell? You would have to > use "eval", like that: > > $ a=hello > $ b=world > $ c='${a} ${b}' > $ a=goodbye > $ b=sunshine > $ d=$(eval echo $c) > $ echo $d > goodbye sunshine > > But please note that "eval" is _not_ recursive!! > > $ a='$b' > $ eval echo $c > $b sunshine > > And this is why I object against this patch. > > Oh, and in U-Boot you could write this as: > > => setenv a hello > => setenv b world > => setenv c '${a} ${b}' > => setenv a goodbye > => setenv b sunshine > => setenv foo "setenv d ${c}! ${a}..." > => run foo > => printenv d > d=goodbye sunshine! goodbye... > > And yes, here you have to be careful about using ' or " as there is > no recursion like you might expect. > > So yes, it would be nice if we had "eval" (which will ocme with the > hush update), and no, "eval" does not recurse either. > > > > Best regards, > > Wolfgang Denk > > -- > DENX Software Engineering GmbH, Managing Director: Wolfgang Denk > HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany > Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de > Real computer scientists despise the idea of actual hardware. Hard- > ware has limitations, software doesn't. It's a real shame that Turing > machines are so poor at I/O.